PostHog vs Amplitude

Agent access

1. ai-native userPoint an agent at llms.txt or agent-oriented docs

   weight 2 · round drawn
   PostHog✓fullprobed9/10

   PostHog serves a working llms.txt (probe confirms HTTP 200 with structured content describing the platform) and extensive agent-oriented docs like the MCP integration page (posthog-docs-12, 19, 24, 31) explicitly designed for AI agents/editors to consume. Individual docs pages also expose `.md` variants as seen in citations (e.g., product-analytics.md, session-replay.md), showing a broader agent-readable docs pattern. Missing for 10: a root docs.md aggregate returned 404, so machine-readable coverage isn't perfectly uniform across all docs.

   • [probe] “PROBE llms.txt: HTTP 200 at https://posthog.com/llms.txt # PostHog > PostHog is the platform for self-driving products. It pairs the full c…”
   • [claimed-docs] “"Show me the full stack trace for the most recent crash, then propose a fix." – Pulls the stack trace, error message, and metadata so the ag…”
   • [claimed-docs] “with just plain text questions your agents can ship a feature flag from a prompt, dig into a stack trace without leaving your editor, run a …”
   • [claimed-docs] “your agents can ship a feature flag from a prompt, dig into a stack trace without leaving your editor, run a HogQL query through Claude”
   • [probe] “official MCP server documented at https://posthog.com/docs/model-context-protocol”
   • [probe] “PROBE docs-md: HTTP 404 at https://posthog.com/docs.md”

   proof ↗⚑ flag

   Amplitude✓fullprobed9/10

   Direct probe confirms a live llms.txt at amplitude.com/llms.txt describing itself as an AI-native platform, plus a docs.md machine-readable docs mirror and an OpenAPI spec, giving agents multiple structured entry points. missing for 10: independent third-party confirmation that agents actually consume these files effectively in practice.

   • [probe] “PROBE llms.txt: HTTP 200 at https://amplitude.com/llms.txt # Amplitude Amplitude is the AI-native digital analytics platform built for AI b…”
   • [probe] “PROBE docs-md: HTTP 200 at https://amplitude.com/docs.md # Amplitude Documentation Official documentation for Amplitude analytics, SDKs, an…”
   • [probe] “PROBE openapi: HTTP 200 at https://amplitude.com/openapi.json — contains "openapi" key”

   proof ↗⚑ flag

2. ai-native userRun the product headlessly / in CI for automation

   weight 2 · round to PostHog
   PostHog✓fullprobed7/10

   PostHog's CLI is explicitly documented for use in 'local scripts, and CI/CD pipelines' alongside terminal and coding agent use, plus REST/HTTP APIs and OpenAPI spec enable headless automation of data capture, queries, and flag management. missing for 10: independent/hands-on confirmation of CI/CD usage, and no explicit CI example/config shown (e.g., GitHub Actions snippet).

   • [claimed-docs] “The PostHog CLI lets you use PostHog from your terminal, your coding agents, local scripts, and CI/CD pipelines.”
   • [probe] “official CLI documented at https://posthog.com/docs/cli”
   • [claimed-docs] “capture events from any language that can send HTTP requests”
   • [probe] “PROBE openapi: HTTP 200 at https://posthog.com/openapi.json — contains "openapi" key”
   • [claimed-docs] “Inject and upload source maps for PostHog error tracking”

   proof ↗⚑ flag

   Amplitude~partialprobed6/10

   Amplitude offers headless automation surfaces suitable for CI: HTTP V2 event ingestion API, Dashboard REST API, Experiment APIs for managing flags/experiments 'from CI, scripts, or your own tools', a CLI-based wizard, and an OpenAPI spec probe. These support scripted/automated use without a GUI, but there's no evidence of a full CLI for ongoing operations (only a setup wizard), no CI-specific recipes/examples, and no independent/hands-on confirmation that headless CI workflows work reliably. missing for 10: dedicated CLI beyond setup wizard, documented CI/CD pipeline examples, independent verification of headless automation in practice.

   • [claimed-docs] “Send data directly from your server to the HTTP V2 endpoint.”
   • [claimed-docs] “Manage flags, experiments, and deployments from CI, scripts, or your own tools with the Experiment APIs.”
   • [claimed-docs] “Run the Wizard CLI to instrument your codebase and get a first dashboard from one terminal command.”
   • [claimed-docs] “Get data that's displayed on the dashboard graphs in JSON format through the Dashboard REST API.”
   • [probe] “PROBE openapi: HTTP 200 at https://amplitude.com/openapi.json — contains "openapi" key”

   proof ↗⚑ flag

3. ai-native userPlug MCP servers into this product so it can use their tools

   weight 3 · round drawn
   PostHog—none0/10

   All MCP-related evidence describes PostHog exposing its own MCP server so external AI clients/editors (Claude, Cursor, etc.) can query PostHog's data and tools — the reverse direction of this story. There is no evidence that PostHog itself acts as an MCP client that can plug in external MCP servers to use their tools.

   • [claimed-docs] “Run trends, funnels, retention, and SQL queries from any MCP client or AI editor.”
   • [claimed-docs] “"Show me the full stack trace for the most recent crash, then propose a fix." – Pulls the stack trace, error message, and metadata so the ag…”
   • [claimed-docs] “with just plain text questions your agents can ship a feature flag from a prompt, dig into a stack trace without leaving your editor, run a …”
   • [claimed-docs] “your agents can ship a feature flag from a prompt, dig into a stack trace without leaving your editor, run a HogQL query through Claude”
   • [claimed-docs] “with just plain text questions your agents can ship a feature flag from a prompt, dig into a stack trace without leaving your editor”
   • [probe] “official MCP server documented at https://posthog.com/docs/model-context-protocol”
   • [github] “You can steer it all from Slack, web, desktop (PostHog Desktop), or your own editor via the MCP.”

   proof ↗⚑ flag

   Amplitude—none0/10

   All evidence describes Amplitude publishing its own MCP server so external AI tools (Claude, Cursor, etc.) can pull Amplitude data — this is Amplitude acting as an MCP *server*, not as a client that plugs in and consumes external MCP servers' tools. There is no evidence Amplitude itself connects to or invokes third-party MCP servers to extend its own AI features.

   • [claimed-docs] “Bring Amplitude’s behavioral context directly into your AI tools—Claude, Cursor, OpenAI, Lovable, Figma, and more—to get answers where you w…”
   • [claimed-docs] “Bring Amplitude's behavioral context directly into your AI tools—Claude, Cursor, OpenAI, Lovable, Figma, and more—to get answers where you w…”
   • [claimed-docs] “Start a conversation about your users or product and get expert-level answers in seconds that you can share with your whole org. No SQL, con…”
   • [probe] “official MCP server documented at https://amplitude.com/mcp-server”

   proof ↗⚑ flag

4. ai-native userConnect an agent via an official MCP server

   weight 3 · round drawn
   PostHog✓fullprobed8/10

   PostHog documents an official MCP server (docs/model-context-protocol) enabling agents to run trends, funnels, HogQL queries, ship feature flags, and pull stack traces directly from MCP clients like Claude, corroborated by the GitHub README describing steering via MCP and a dedicated probe confirming the docs page exists. Missing for 10: independent third-party hands-on validation of the MCP server (only vendor docs/GitHub and one unrelated community complaint about a different AI assistant feature).

   • [claimed-docs] “"Show me the full stack trace for the most recent crash, then propose a fix." – Pulls the stack trace, error message, and metadata so the ag…”
   • [claimed-docs] “with just plain text questions your agents can ship a feature flag from a prompt, dig into a stack trace without leaving your editor, run a …”
   • [claimed-docs] “your agents can ship a feature flag from a prompt, dig into a stack trace without leaving your editor, run a HogQL query through Claude”
   • [claimed-docs] “with just plain text questions your agents can ship a feature flag from a prompt, dig into a stack trace without leaving your editor”
   • [github] “You can steer it all from Slack, web, desktop (PostHog Desktop), or your own editor via the MCP.”
   • [probe] “official MCP server documented at https://posthog.com/docs/model-context-protocol”

   proof ↗⚑ flag

   Amplitude✓fullprobed8/10

   Amplitude is not itself an agent but a SaaS analytics platform, so the axis applies; it documents an official MCP server at amplitude.com/mcp-server that brings Amplitude's behavioral data into AI tools like Claude, Cursor, OpenAI, and Lovable, corroborated by a probe confirming the page exists. Missing for 10: independent/hands-on third-party verification of the MCP server actually working in practice.

   • [claimed-docs] “Bring Amplitude’s behavioral context directly into your AI tools—Claude, Cursor, OpenAI, Lovable, Figma, and more—to get answers where you w…”
   • [claimed-docs] “Bring Amplitude's behavioral context directly into your AI tools—Claude, Cursor, OpenAI, Lovable, Figma, and more—to get answers where you w…”
   • [claimed-docs] “Start a conversation about your users or product and get expert-level answers in seconds that you can share with your whole org. No SQL, con…”
   • [probe] “official MCP server documented at https://amplitude.com/mcp-server”

   proof ↗⚑ flag

5. ai-native userUse an official CLI

   weight 2 · round to PostHog
   PostHog✓fullprobed8/10

   PostHog documents an official CLI usable from terminal, coding agents, local scripts, and CI/CD pipelines, with concrete features like source-map injection/upload for error tracking, and this is corroborated by a dedicated docs page and a probe confirming the CLI documentation is live. missing for 10: independent hands-on community review of the CLI itself (community evidence only covers MCP/LLM features, not the CLI specifically).

   • [claimed-docs] “The PostHog CLI lets you use PostHog from your terminal, your coding agents, local scripts, and CI/CD pipelines.”
   • [claimed-docs] “Inject and upload source maps for PostHog error tracking”
   • [probe] “official CLI documented at https://posthog.com/docs/cli”

   proof ↗⚑ flag

   Amplitude~partialclaimed5/10

   Amplitude ships a Wizard CLI that auto-instruments a codebase and gets a first dashboard from one terminal command (amplitude-docs-27, amplitude-docs-1), which counts as an official CLI, but this is scoped to onboarding/setup rather than a general-purpose AI-native CLI for ongoing agentic workflows. Missing for 10: documentation of a broader CLI with commands beyond initial instrumentation, evidence of AI-native/agentic use cases via the CLI, and independent hands-on corroboration of the CLI's capabilities.

   • [claimed-docs] “Run the Wizard CLI to instrument your codebase and get a first dashboard from one terminal command.”
   • [claimed-docs] “The wizard detects your framework, installs the right SDK, adds your API key, and fires a test event.”

   proof ↗⚑ flag

6. ai-native userDrive the product through a documented public API

   weight 3 · round to PostHog
   PostHog✓fullprobed9/10

   PostHog has a documented public API (posthog.com/docs/api.md, HTTP capture from any language), a live OpenAPI spec confirmed by probe (openapi.json), an official CLI for terminal/CI/CD use, and a documented MCP server enabling AI agents to run queries, ship flags, and inspect stack traces via plain text — all first-party documented and probe-verified. Missing for 10: independent hands-on corroboration of the API/OpenAPI usage beyond vendor docs and probes.

   • [claimed-docs] “capture events from any language that can send HTTP requests”
   • [probe] “PROBE openapi: HTTP 200 at https://posthog.com/openapi.json — contains "openapi" key”
   • [probe] “official MCP server documented at https://posthog.com/docs/model-context-protocol”
   • [probe] “official CLI documented at https://posthog.com/docs/cli”
   • [claimed-docs] “with just plain text questions your agents can ship a feature flag from a prompt, dig into a stack trace without leaving your editor, run a …”
   • [claimed-docs] “The PostHog CLI lets you use PostHog from your terminal, your coding agents, local scripts, and CI/CD pipelines.”
   • [claimed-docs] “"Show me the full stack trace for the most recent crash, then propose a fix." – Pulls the stack trace, error message, and metadata so the ag…”

   proof ↗⚑ flag

   Amplitude✓fullprobed8/10

   Amplitude documents a broad public HTTP API surface (event ingestion, dashboard REST API, cohort management, event/property CRUD, experiment/feature-flag APIs) plus a confirmed OpenAPI spec and llms.txt/docs.md endpoints for machine consumption, indicating a well-documented, programmatically drivable API. missing for 10: independent third-party corroboration of API completeness/reliability and no evidence of official client libraries per language beyond SDK mentions.

   • [claimed-docs] “Send data directly from your server to the HTTP V2 endpoint.”
   • [claimed-docs] “Export your project's event data.”
   • [claimed-docs] “Manage flags, experiments, and deployments from CI, scripts, or your own tools with the Experiment APIs.”
   • [claimed-docs] “List all your cohorts in Amplitude, export a cohort in Amplitude, or upload a cohort.”
   • [claimed-docs] “Get data that's displayed on the dashboard graphs in JSON format through the Dashboard REST API.”
   • [claimed-docs] “Create, get, update, and delete categories, event types, event properties, and user properties.”
   • [probe] “PROBE openapi: HTTP 200 at https://amplitude.com/openapi.json — contains "openapi" key”
   • [probe] “PROBE llms.txt: HTTP 200 at https://amplitude.com/llms.txt # Amplitude Amplitude is the AI-native digital analytics platform built for AI b…”
   • [probe] “PROBE docs-md: HTTP 200 at https://amplitude.com/docs.md # Amplitude Documentation Official documentation for Amplitude analytics, SDKs, an…”

   proof ↗⚑ flag

7. ai-native userIssue scoped/least-privilege API credentials for an agent

   weight 2 · round drawn
   PostHog—none0/10

   No evidence describes scoped or least-privilege API key/credential issuance for agents — the API docs mention basic HTTP event capture and an OpenAPI spec exists, but nothing about permission scoping, token minting with restricted access, or credential management for AI agents specifically.

   ⚑ flag

   Amplitude—none0/10

   Evidence shows API keys, an MCP server, and OpenAPI spec, but nothing about scoped/least-privilege credential issuance (e.g., role-based API tokens, granular permission scopes) for agent use is documented.

   ⚑ flag

8. ai-native userBuild against official SDKs

   weight 2 · round to Amplitude
   PostHog✓fullprobed7/10

   PostHog documents official SDKs/APIs (HTTP capture from any language, OpenAPI spec, CLI, MCP server) that let AI-native builders integrate programmatically, including AI-specific tooling like LLM observability SDKs and MCP-based agent workflows (posthog-docs-11, posthog-probe-3, posthog-docs-20/24, posthog-gh-6/9). A hands-on community comment corroborates smooth SDK integration for a Swift iOS app (posthog-comm-3), though another notes friction with LLM/AI query features (posthog-comm-8), suggesting some rough edges. Missing for 10: an explicit enumerated list of officially maintained language SDKs and stronger independent third-party validation of SDK completeness/versioning.

   • [claimed-docs] “capture events from any language that can send HTTP requests”
   • [probe] “PROBE openapi: HTTP 200 at https://posthog.com/openapi.json — contains "openapi" key”
   • [claimed-docs] “The PostHog CLI lets you use PostHog from your terminal, your coding agents, local scripts, and CI/CD pipelines.”
   • [claimed-docs] “your agents can ship a feature flag from a prompt, dig into a stack trace without leaving your editor, run a HogQL query through Claude”
   • [github] “Capture traces, generations, latency, and cost for your LLM-powered app.”
   • [github] “AI observability: Capture traces, generations, latency, and cost for your LLM-powered app.”
   • [community] “I've been using PostHog with my app for about a week now, and so far the results have been good. Pretty straightforward to integrate with a …”
   • [community] “I really wish you guys would do something about the llm integration, I tried using it two days ago to create a cohort of users using a sql q…”

   proof ↗⚑ flag

   Amplitude✓fullprobed8/10

   Amplitude documents a full suite of official SDKs (web, mobile, server, game engines) sharing a unified identity/event model and API surface, backed by an installer wizard, HTTP V2 ingestion endpoint, and a published OpenAPI spec, all discoverable via llms.txt/docs.md for AI-native consumption. Missing for 10: independent/hands-on developer corroboration of SDK integration ease and no explicit AI-agent-specific SDK usage examples beyond the MCP server.

   • [claimed-docs] “Evaluate flags locally for low-latency rollout decisions.”
   • [claimed-docs] “Auto-instrument page views, clicks, and sessions out of the box.”
   • [claimed-docs] “Every SDK shares the same identity, event, and consent model, so a user instrumented once is understood everywhere.”
   • [claimed-docs] “Run the Wizard CLI to instrument your codebase and get a first dashboard from one terminal command.”
   • [claimed-docs] “Use Experiment SDKs to flag and test features across web, mobile, and backend.”
   • [claimed-docs] “Use the same API surface across web, mobile, server, and game engines.”
   • [claimed-docs] “Unified SDKs provide one integration point for multiple Amplitude products.”
   • [claimed-docs] “Send data directly from your server to the HTTP V2 endpoint.”
   • [probe] “PROBE llms.txt: HTTP 200 at https://amplitude.com/llms.txt # Amplitude Amplitude is the AI-native digital analytics platform built for AI b…”
   • [probe] “PROBE docs-md: HTTP 200 at https://amplitude.com/docs.md # Amplitude Documentation Official documentation for Amplitude analytics, SDKs, an…”
   • [probe] “PROBE openapi: HTTP 200 at https://amplitude.com/openapi.json — contains "openapi" key”

   proof ↗⚑ flag

9. ai-native userSubscribe to events via webhooks

   weight 2 · round to PostHog
   PostHog~partialclaimed4/10

   PostHog's CDP pipelines can send event data to Slack, PagerDuty, and other SaaS destinations, and can filter/reshape events and route them 'wherever it needs to go' in realtime, which implies webhook-style outbound delivery — but no evidence pack item explicitly names 'webhook' as a subscription mechanism or ties this to AI-agent consumption of the event stream. missing for 10: explicit webhook endpoint/subscription documentation, confirmation that arbitrary event streams can be pushed to a caller-defined webhook URL, and any AI-native usage example of subscribing via webhook.

   • [claimed-docs] “It pulls data in from the tools you already use, lets you filter and reshape events as they're ingested, and sends the result wherever it ne…”
   • [claimed-docs] “It pulls data in from the tools you already use, lets you filter and reshape events as they're ingested, and sends the result wherever it ne…”
   • [claimed-docs] “Slack: Send events to Slack channels SaaS tools: Send events to SaaS tools like Braze, Customer.io, and more”
   • [claimed-docs] “Send alerts to Slack, PagerDuty, or monitoring tools”
   • [github] “Create workflows that automate actions or send messages to your users.”

   proof ↗⚑ flag

   Amplitude—none0/10

   No evidence of any webhook subscription mechanism for events; Amplitude's evidence covers SDKs, APIs for ingest/export, integrations, and an MCP server, but nothing describes outbound webhooks for event subscriptions. missing for 10: any webhook API/documentation, event subscription mechanism, or third-party confirmation of webhook support.

   ⚑ flag

Agentic features

1. ai-native userGet AI-generated insights and suggestions from my data inside the product

   weight 2 · round to Amplitude
   PostHog!disputedcontradicted5/10

   PostHog markets in-product AI ('PostHog AI', 'self-driving products') that can suggest code fixes from stack traces, auto-write Hog functions, and turn product signals into researched reports/PRs (posthog-docs-12, posthog-docs-28, posthog-gh-1, posthog-probe-1). However a hands-on community report describes the in-app LLM feature failing a basic task (creating a SQL-based cohort) and severely degrading device performance, directly contradicting the polished 'insights on demand' claim (posthog-comm-8). missing for 10: independent corroboration that AI-generated insights/suggestions work reliably at scale, no rebuttal or fix documented for the reported failure.

   • [claimed-docs] “"Show me the full stack trace for the most recent crash, then propose a fix." – Pulls the stack trace, error message, and metadata so the ag…”
   • [claimed-docs] “Pipelines are built from Hog functions, which you can configure from a template, write yourself, or have PostHog AI write for you.”
   • [github] “Turn signals in your product data (errors, rage clicks, failed queries, and more) into researched reports and pull requests you review and m…”
   • [probe] “PROBE llms.txt: HTTP 200 at https://posthog.com/llms.txt # PostHog > PostHog is the platform for self-driving products. It pairs the full c…”
   • [community] “I really wish you guys would do something about the llm integration, I tried using it two days ago to create a cohort of users using a sql q…”

   proof ↗⚑ flag

   Amplitude✓fullprobed8/10

   Amplitude AI (docs-10, docs-34) explicitly lets users ask plain-language questions and get expert-level answers/insights across their data, and the MCP server (docs-11/17, probe-4) surfaces this behavioral context directly inside AI tools. This directly matches the story of getting AI-generated insights inside the product. Missing for 10: independent/hands-on validation of the AI insight quality (only vendor docs present), and no detail on proactive suggestion generation vs. query-response only.

   • [claimed-docs] “Ask questions in plain language — Amplitude AI surfaces answers across all your data.”
   • [claimed-docs] “Start a conversation about your users or product and get expert-level answers in seconds that you can share with your whole org. No SQL, con…”
   • [claimed-docs] “Bring Amplitude’s behavioral context directly into your AI tools—Claude, Cursor, OpenAI, Lovable, Figma, and more—to get answers where you w…”
   • [claimed-docs] “Bring Amplitude's behavioral context directly into your AI tools—Claude, Cursor, OpenAI, Lovable, Figma, and more—to get answers where you w…”
   • [probe] “official MCP server documented at https://amplitude.com/mcp-server”

   proof ↗⚑ flag

2. ai-native userSet up automations that run autonomously in the background

   weight 2 · round to PostHog
   PostHog~partialcommunity6/10

   PostHog's CDP pipelines and workflows can run automatically in realtime or on a schedule (posthog-docs-18, posthog-gh-7), with Hog functions optionally written by PostHog AI (posthog-docs-28), and alerts/feature-flag automation exist (posthog-docs-3, posthog-docs-5). However, evidence centers on data-pipeline/workflow automation rather than an AI-native agent explicitly setting up and running autonomous background jobs end-to-end, and one community report notes friction with AI-driven query/cohort creation (posthog-comm-8). Missing for 10: first-party documentation of an AI agent configuring and monitoring a fully autonomous background automation, independent success reports of such automations running unattended.

   • [claimed-docs] “It pulls data in from the tools you already use, lets you filter and reshape events as they're ingested, and sends the result wherever it ne…”
   • [github] “Create workflows that automate actions or send messages to your users.”
   • [claimed-docs] “Pipelines are built from Hog functions, which you can configure from a template, write yourself, or have PostHog AI write for you.”
   • [claimed-docs] “Build insights, assemble dashboards, share them, and set alerts on the numbers that matter.”
   • [community] “I really wish you guys would do something about the llm integration, I tried using it two days ago to create a cohort of users using a sql q…”

   proof ↗⚑ flag

   Amplitude~partialclaimed4/10

   Amplitude offers pieces that resemble autonomous background automation—flag rollout/rollback rules that react to data (amplitude-docs-19), automated in-product guides/surveys triggers (amplitude-docs-14), CI/script-driven experiment management (amplitude-docs-13), and an embeddable agent that 'completes tasks for users' (amplitude-docs-26)—but there is no dedicated, documented automation/workflow builder explicitly for AI-native users to configure and run persistent background automations. missing for 10: a first-party 'automations' or workflow engine with triggers/schedules, independent evidence of autonomous background execution, and confirmation the embedded agent runs unattended rather than on-demand.

   • [claimed-docs] “Release behind a flag, ramp to the audience you choose, and turn it off the moment something looks wrong.”
   • [claimed-docs] “Launch in-product guides and surveys to onboard, announce, and learn from users.”
   • [claimed-docs] “Manage flags, experiments, and deployments from CI, scripts, or your own tools with the Experiment APIs.”
   • [claimed-docs] “Embed an agent grounded in your help content and user behavior. Answer questions, guide flows, and complete tasks for users in the moments t…”

   proof ↗⚑ flag

3. ai-native userDelegate tasks to a built-in AI assistant inside the product

   weight 3 · round to Amplitude
   PostHog!disputedcontradicted4/10

   PostHog documents an in-product AI assistant/agent ("PostHog AI") that can write Hog functions, turn product signals into researched reports and PRs, and be steered from Slack/web/desktop/editor (posthog-docs-28, posthog-gh-1, posthog-gh-2), which matches the story of delegating tasks to a built-in assistant. However, a hands-on community report describes the LLM integration failing to perform a requested task (creating a cohort via SQL) and causing severe performance degradation, directly contradicting smooth delegation (posthog-comm-8). Missing for 10: first-party detail on the assistant's task scope/reliability, more independent corroboration of successful delegation, and resolution of the reported failure.

   • [claimed-docs] “Pipelines are built from Hog functions, which you can configure from a template, write yourself, or have PostHog AI write for you.”
   • [github] “Turn signals in your product data (errors, rage clicks, failed queries, and more) into researched reports and pull requests you review and m…”
   • [github] “You can steer it all from Slack, web, desktop (PostHog Desktop), or your own editor via the MCP.”
   • [community] “I really wish you guys would do something about the llm integration, I tried using it two days ago to create a cohort of users using a sql q…”

   proof ↗⚑ flag

   Amplitude~partialclaimed6/10

   Amplitude AI lets users ask plain-language questions and get answers across their data, and an embeddable agent can 'answer questions, guide flows, and complete tasks for users,' which shows some built-in delegatable AI assistant capability. However, evidence is limited to marketing docs snippets with no hands-on demonstration of task delegation scope, autonomy limits, or independent corroboration. missing for 10: independent/hands-on verification of the AI assistant actually completing delegated tasks, detail on task scope/autonomy, and confirmation this is available to the analytics-user persona (not just embedded end-user product).

   • [claimed-docs] “Ask questions in plain language — Amplitude AI surfaces answers across all your data.”
   • [claimed-docs] “Embed an agent grounded in your help content and user behavior. Answer questions, guide flows, and complete tasks for users in the moments t…”
   • [claimed-docs] “Start a conversation about your users or product and get expert-level answers in seconds that you can share with your whole org. No SQL, con…”

   proof ↗⚑ flag

4. ai-native userOperate the product with natural-language commands

   weight 2 · round to Amplitude
   PostHog!disputedcontradicted5/10

   PostHog documents natural-language operation via its MCP integration—shipping feature flags, running HogQL queries, and debugging stack traces 'from a prompt' or 'plain text questions' (posthog-docs-12, posthog-docs-19, posthog-docs-24, posthog-docs-31, posthog-probe-4). However, a hands-on community report directly contradicts this: a user tried the LLM integration to create a cohort via SQL and it failed with an incorrect response, plus caused severe performance degradation (posthog-comm-8). missing for 10: independent corroboration that the failure was fixed, broader hands-on validation of natural-language commands across other product areas (analytics, experiments, session replay) beyond MCP docs.

   • [claimed-docs] “"Show me the full stack trace for the most recent crash, then propose a fix." – Pulls the stack trace, error message, and metadata so the ag…”
   • [claimed-docs] “with just plain text questions your agents can ship a feature flag from a prompt, dig into a stack trace without leaving your editor, run a …”
   • [claimed-docs] “your agents can ship a feature flag from a prompt, dig into a stack trace without leaving your editor, run a HogQL query through Claude”
   • [claimed-docs] “with just plain text questions your agents can ship a feature flag from a prompt, dig into a stack trace without leaving your editor”
   • [probe] “official MCP server documented at https://posthog.com/docs/model-context-protocol”
   • [community] “I really wish you guys would do something about the llm integration, I tried using it two days ago to create a cohort of users using a sql q…”

   proof ↗⚑ flag

   Amplitude✓fullprobed8/10

   Amplitude AI lets users 'ask questions in plain language' and get answers across data, and the MCP server brings this natural-language querying into external AI tools (Claude, Cursor, etc.), directly enabling natural-language operation of the product's analytics capabilities. This is corroborated by dedicated docs pages and a probe confirming the MCP server exists. Missing for 10: independent/hands-on verification of natural-language command accuracy and breadth (e.g., can it also configure flags/dashboards via NL, not just query), and no community evidence testing this specific feature.

   • [claimed-docs] “Ask questions in plain language — Amplitude AI surfaces answers across all your data.”
   • [claimed-docs] “Bring Amplitude’s behavioral context directly into your AI tools—Claude, Cursor, OpenAI, Lovable, Figma, and more—to get answers where you w…”
   • [claimed-docs] “Bring Amplitude's behavioral context directly into your AI tools—Claude, Cursor, OpenAI, Lovable, Figma, and more—to get answers where you w…”
   • [claimed-docs] “Start a conversation about your users or product and get expert-level answers in seconds that you can share with your whole org. No SQL, con…”
   • [probe] “official MCP server documented at https://amplitude.com/mcp-server”

   proof ↗⚑ flag

Api quality

1. ai-native userExplore an interactive API reference with runnable examples

   weight 2 · round to PostHog
   PostHog~partialprobed3/10

   The evidence shows PostHog publishes a machine-readable OpenAPI spec (openapi.json) and API docs referencing HTTP-based event capture, which implies some API reference exists, but there is no direct evidence of an interactive reference page with runnable/try-it examples. missing for 10: confirmation of an interactive API console, runnable code snippets, or live 'try it' functionality in the docs.

   • [probe] “PROBE openapi: HTTP 200 at https://posthog.com/openapi.json — contains "openapi" key”
   • [claimed-docs] “capture events from any language that can send HTTP requests”

   proof ↗⚑ flag

   Amplitude—none0/10

   Evidence confirms Amplitude has API documentation (HTTP V2 endpoint, Dashboard REST API, various CRUD endpoints) and even an openapi.json spec file, but nothing in the pack shows an interactive reference UI with runnable/try-it examples — no Swagger-style console, no code sandbox, no mention of 'try it' functionality.

   • [claimed-docs] “Send data directly from your server to the HTTP V2 endpoint.”
   • [claimed-docs] “Get data that's displayed on the dashboard graphs in JSON format through the Dashboard REST API.”
   • [claimed-docs] “Create, get, update, and delete categories, event types, event properties, and user properties.”
   • [probe] “PROBE openapi: HTTP 200 at https://amplitude.com/openapi.json — contains "openapi" key”

   proof ↗⚑ flag

2. ai-native userDownload a machine-readable API spec (OpenAPI or equivalent)

   weight 2 · round drawn
   PostHog✓fullprobed9/10

   A probe confirms PostHog serves a machine-readable OpenAPI spec at https://posthog.com/openapi.json (HTTP 200, contains 'openapi' key), and the API docs reference programmatic access via HTTP. Missing for 10: no explicit first-party docs page describing the OpenAPI spec's usage/versioning or independent community confirmation of its completeness.

   • [probe] “PROBE openapi: HTTP 200 at https://posthog.com/openapi.json — contains "openapi" key”
   • [claimed-docs] “capture events from any language that can send HTTP requests”

   proof ↗⚑ flag

   Amplitude✓fullprobed9/10

   A direct probe confirms a machine-readable OpenAPI spec is available at https://amplitude.com/openapi.json (contains an 'openapi' key), and Amplitude also exposes llms.txt/docs.md machine-readable docs. Missing for 10: no independent/third-party corroboration of the spec's completeness or usage examples beyond the probe.

   • [probe] “PROBE openapi: HTTP 200 at https://amplitude.com/openapi.json — contains "openapi" key”
   • [probe] “PROBE llms.txt: HTTP 200 at https://amplitude.com/llms.txt # Amplitude Amplitude is the AI-native digital analytics platform built for AI b…”
   • [probe] “PROBE docs-md: HTTP 200 at https://amplitude.com/docs.md # Amplitude Documentation Official documentation for Amplitude analytics, SDKs, an…”
   • [claimed-docs] “Send data directly from your server to the HTTP V2 endpoint.”

   proof ↗⚑ flag

3. ai-native userTest against a sandbox environment without touching production data

   weight 1 · round drawn
   PostHog—none0/10

   The evidence pack covers self-hosting (Docker Compose/MIT license), MCP integration, feature flags, and experiments, but nothing describes a dedicated sandbox/staging environment or test-mode data isolation from production. Self-hosting is a deployment option, not a sandbox testing feature, so it does not satisfy this story.

   ⚑ flag

   Amplitude—none0/10

   No evidence pack mentions a sandbox, test project, or staging environment separate from production data; docs reference API keys, SDKs, and CI/deployment but nothing about isolating test data from production. Missing for 10: sandbox/test environment documentation, guidance on separating test vs production data, any staging API keys or environment flags.

   ⚑ flag

4. ai-native userRely on versioned APIs with a documented deprecation policy

   weight 2 · round drawn
   PostHog—none0/10

   There is an OpenAPI spec probe (posthog-probe-3) confirming an API exists, but no evidence anywhere in the pack of API versioning scheme or a documented deprecation policy for that API.

   • [probe] “PROBE openapi: HTTP 200 at https://posthog.com/openapi.json — contains "openapi" key”

   proof ↗⚑ flag

   Amplitude—none0/10

   Evidence shows Amplitude has REST/HTTP APIs (e.g., HTTP V2 endpoint, Dashboard REST API, Experiment APIs) and an OpenAPI spec, confirming APIs exist, but nothing in the pack documents API versioning conventions or a deprecation/sunset policy for these APIs. The axis applies to any API-driven platform, but no evidence supports the specific claim of a documented deprecation policy.

   • [claimed-docs] “Send data directly from your server to the HTTP V2 endpoint.”
   • [claimed-docs] “Get data that's displayed on the dashboard graphs in JSON format through the Dashboard REST API.”
   • [claimed-docs] “Manage flags, experiments, and deployments from CI, scripts, or your own tools with the Experiment APIs.”
   • [probe] “PROBE openapi: HTTP 200 at https://amplitude.com/openapi.json — contains "openapi" key”

   proof ↗⚑ flag

1. ai-native userPerform bulk operations across many items at once

   weight 2 · round to Amplitude
   PostHog~partialcommunity4/10

   PostHog offers SQL for querying/modifying data at scale, a CDP pipeline that can reshape/filter events in bulk, and a CLI/API for scripting operations, which could underpin bulk actions, but there is no explicit documentation of a 'bulk update/delete/tag N items' feature, and a hands-on report shows the AI/SQL interface failing on a basic bulk task (creating a cohort via SQL). missing for 10: explicit bulk-operation API/UI (e.g., mass edit flags, batch delete events), documented bulk-scale limits, and independent confirmation that SQL/CLI bulk actions work reliably for AI-driven workflows.

   • [claimed-docs] “You can use SQL (Structured Query Language) throughout PostHog to manage, query, and modify data.”
   • [claimed-docs] “Sync Stripe, Postgres, Salesforce, HubSpot, and dozens of other sources, then query them alongside your events, persons, and sessions in one…”
   • [claimed-docs] “It pulls data in from the tools you already use, lets you filter and reshape events as they're ingested, and sends the result wherever it ne…”
   • [claimed-docs] “The PostHog CLI lets you use PostHog from your terminal, your coding agents, local scripts, and CI/CD pipelines.”
   • [community] “I really wish you guys would do something about the llm integration, I tried using it two days ago to create a cohort of users using a sql q…”

   proof ↗⚑ flag

   Amplitude~partialclaimed5/10

   Amplitude's APIs support programmatic mass operations—batch event ingestion via HTTP V2 (docs-2), CRUD across event types/properties (docs-38), cohort list/export/upload (docs-29), and CI/script-driven flag/experiment management (docs-13)—which enable bulk-style automation for AI-native users. However there's no explicit documentation of a dedicated bulk-operations UI or API endpoint for acting on 'many items at once' as a single call. Missing for 10: explicit bulk-update/bulk-delete endpoints, documented batch size limits, and independent evidence of bulk workflows being used successfully.

   • [claimed-docs] “Send data directly from your server to the HTTP V2 endpoint.”
   • [claimed-docs] “Create, get, update, and delete categories, event types, event properties, and user properties.”
   • [claimed-docs] “List all your cohorts in Amplitude, export a cohort in Amplitude, or upload a cohort.”
   • [claimed-docs] “Manage flags, experiments, and deployments from CI, scripts, or your own tools with the Experiment APIs.”
   • [claimed-docs] “Export your project's event data.”

   proof ↗⚑ flag

2. ai-native userDefine rules that trigger actions automatically on events

   weight 3 · round to PostHog
   PostHog✓fullclaimed7/10

   PostHog's CDP/Pipelines ('destinations') and alerting system let users define trigger conditions on events (errors, rage clicks, property changes, etc.) that automatically fire actions like Slack alerts, CRM syncs, PagerDuty notifications, or custom Hog functions, and workflows can automate actions/messages to users. missing for 10: no independent/hands-on verification of complex multi-condition automation rules, and no detail on latency/reliability of the automation engine at scale.

   • [claimed-docs] “It pulls data in from the tools you already use, lets you filter and reshape events as they're ingested, and sends the result wherever it ne…”
   • [claimed-docs] “It pulls data in from the tools you already use, lets you filter and reshape events as they're ingested, and sends the result wherever it ne…”
   • [claimed-docs] “Pipelines are built from Hog functions, which you can configure from a template, write yourself, or have PostHog AI write for you.”
   • [claimed-docs] “Slack: Send events to Slack channels SaaS tools: Send events to SaaS tools like Braze, Customer.io, and more”
   • [claimed-docs] “Send alerts to Slack, PagerDuty, or monitoring tools”
   • [github] “Create workflows that automate actions or send messages to your users.”

   proof ↗⚑ flag

   Amplitude—none0/10

   The evidence pack covers analytics, experimentation, session replay, and AI-query features, but contains no documentation of a rules/automation engine that triggers actions automatically based on events (e.g., event-triggered webhooks, workflows, or conditional messaging). Feature flags/experiments and in-product guides are mentioned but only as manually configured campaigns, not as an automatic event-triggered rules system.

   • [claimed-docs] “Launch in-product guides and surveys to onboard, announce, and learn from users.”
   • [claimed-docs] “Release behind a flag, ramp to the audience you choose, and turn it off the moment something looks wrong.”
   • [claimed-docs] “Roll out behind a flag and run A/B tests so the data, not a hunch, drives the release.”

   proof ↗⚑ flag

3. ai-native userSchedule recurring jobs or workflows

   weight 2 · round to PostHog
   PostHog~partialclaimed5/10

   PostHog's CDP/pipelines can run 'in realtime or on a schedule' (posthog-docs-18) and GitHub notes 'workflows that automate actions or send messages to your users' (posthog-gh-7), suggesting some recurring automation capability, but there's no documentation of a general-purpose recurring job/workflow scheduler accessible via API, CLI, or MCP for AI-native orchestration. Missing for 10: explicit docs on scheduling arbitrary jobs/workflows via API or agent tooling, cron-like configuration details, and independent confirmation that scheduled pipelines work reliably.

   • [claimed-docs] “It pulls data in from the tools you already use, lets you filter and reshape events as they're ingested, and sends the result wherever it ne…”
   • [github] “Create workflows that automate actions or send messages to your users.”
   • [claimed-docs] “Pipelines are built from Hog functions, which you can configure from a template, write yourself, or have PostHog AI write for you.”

   proof ↗⚑ flag

   Amplitude—none0/10

   Amplitude is an analytics/experimentation platform; nothing in the evidence pack references scheduling recurring jobs, workflows, or automated cron-like tasks. There's no mention of scheduled reports, recurring data exports, or workflow automation triggers.

   ⚑ flag

4. ai-native userVersion, review, and roll back my automations

   weight 1 · round to Amplitude
   PostHog—none0/10

   No evidence pack items describe version history, review workflows, or rollback capabilities for automations (feature flags, CDP pipelines, or workflows); feature flags can be toggled off but there's no mention of versioning, diffing, or rollback to a previous version. Missing for 10: version history for flags/pipelines, review/approval workflow, rollback mechanism, audit trail of changes.

   ⚑ flag

   Amplitude~partialclaimed4/10

   Amplitude's feature-flag/experiment system supports rollback-like control ('turn it off the moment something looks wrong') and can be managed via CI/scripts, but there is no evidence of a version history, change review, or audit/rollback log for automations. missing for 10: explicit versioning/change-history for flags or automations, a review/approval workflow, and documented rollback-to-prior-version mechanics.

   • [claimed-docs] “Release behind a flag, ramp to the audience you choose, and turn it off the moment something looks wrong.”
   • [claimed-docs] “Manage flags, experiments, and deployments from CI, scripts, or your own tools with the Experiment APIs.”
   • [claimed-docs] “Roll out behind a flag and run A/B tests so the data, not a hunch, drives the release.”

   proof ↗⚑ flag

Ai analytics ops

1. ai-native userHave an agent build a dashboard of my key metrics end-to-end via the API or MCP server

   weight 3 · round to PostHog
   PostHog~partialprobed6/10

   PostHog has a documented official MCP server and public API (openapi.json) that let an agent run trends, funnels, retention, and SQL queries, ship feature flags, and inspect stack traces from an editor/agent (posthog-docs-2, 12, 19, 24, 31, posthog-probe-4, posthog-probe-3). Dashboard assembly, sharing, and alerting are documented product capabilities (posthog-docs-3, 25), but the evidence never explicitly shows the MCP/API path being used to assemble and publish a full dashboard end-to-end, and a hands-on community report describes the AI/LLM integration failing on a related task (cohort creation) and causing severe UI slowdown (posthog-comm-8), raising doubts about reliability of agent-driven workflows. Missing for 10: explicit documentation/demo of an agent creating and saving a dashboard object (not just insights/queries) via MCP or API, and independent corroboration that this specific end-to-end flow works reliably.

   • [claimed-docs] “Run trends, funnels, retention, and SQL queries from any MCP client or AI editor.”
   • [claimed-docs] “Build insights, assemble dashboards, share them, and set alerts on the numbers that matter.”
   • [claimed-docs] “"Show me the full stack trace for the most recent crash, then propose a fix." – Pulls the stack trace, error message, and metadata so the ag…”
   • [claimed-docs] “with just plain text questions your agents can ship a feature flag from a prompt, dig into a stack trace without leaving your editor, run a …”
   • [claimed-docs] “your agents can ship a feature flag from a prompt, dig into a stack trace without leaving your editor, run a HogQL query through Claude”
   • [claimed-docs] “with just plain text questions your agents can ship a feature flag from a prompt, dig into a stack trace without leaving your editor”
   • [probe] “PROBE openapi: HTTP 200 at https://posthog.com/openapi.json — contains "openapi" key”
   • [probe] “official MCP server documented at https://posthog.com/docs/model-context-protocol”
   • [community] “I really wish you guys would do something about the llm integration, I tried using it two days ago to create a cohort of users using a sql q…”

   proof ↗⚑ flag

   Amplitude~partialprobed5/10

   Amplitude has documented APIs (Dashboard REST API for reading chart data, event/property CRUD, cohort APIs) and an official MCP server that lets AI tools like Claude/Cursor query Amplitude data in plain language, which covers agent-driven analytics querying. However, evidence shows the Dashboard API is read-only (fetching JSON of existing dashboard graphs) and 'Create a dashboard' appears to be a UI action, not an API/MCP-exposed operation, so there's no documented end-to-end agent workflow that programmatically builds a new dashboard from scratch via API or MCP. Missing for 10: an API/MCP endpoint or example specifically for creating/configuring dashboards (not just querying data), and any hands-on/community proof of an agent successfully building a dashboard end-to-end.

   • [claimed-docs] “Bring Amplitude’s behavioral context directly into your AI tools—Claude, Cursor, OpenAI, Lovable, Figma, and more—to get answers where you w…”
   • [claimed-docs] “Bring Amplitude's behavioral context directly into your AI tools—Claude, Cursor, OpenAI, Lovable, Figma, and more—to get answers where you w…”
   • [claimed-docs] “Start a conversation about your users or product and get expert-level answers in seconds that you can share with your whole org. No SQL, con…”
   • [claimed-docs] “Get data that's displayed on the dashboard graphs in JSON format through the Dashboard REST API.”
   • [claimed-docs] “Create a dashboard”
   • [claimed-docs] “Create, get, update, and delete categories, event types, event properties, and user properties.”
   • [probe] “official MCP server documented at https://amplitude.com/mcp-server”

   proof ↗⚑ flag

Dashboards

1. developerCreate and manage dashboards and saved insights programmatically via the API

   weight 2 · round to PostHog
   PostHog~partialprobed4/10

   The evidence confirms PostHog exposes a general REST API (openapi.json spec) and documents building/sharing dashboards and insights with alerts, implying programmatic access is plausible, but no citation explicitly documents dashboard/insight CRUD API endpoints or SDK examples for managing them programmatically. missing for 10: explicit API docs/endpoints for creating/updating dashboards and insights, code examples for programmatic dashboard management, independent confirmation of this specific API surface.

   • [probe] “PROBE openapi: HTTP 200 at https://posthog.com/openapi.json — contains "openapi" key”
   • [claimed-docs] “Build insights, assemble dashboards, share them, and set alerts on the numbers that matter.”
   • [claimed-docs] “Build trends, funnels, retention, paths, stickiness, and lifecycle insights on the events you already send to PostHog, then save them to das…”

   proof ↗⚑ flag

   Amplitude~partialclaimed3/10

   The docs mention a Dashboard REST API that returns JSON data for existing dashboard graphs and a general 'Create a dashboard' doc, plus API endpoints for cohorts and event/property metadata, but none of these clearly describe full CRUD (create/update/delete) for dashboards or 'saved insights' via API—only read/export access is explicitly documented. Missing for 10: explicit API endpoints for creating/updating/deleting dashboards, a documented 'saved insights' API object, and independent confirmation that these operations work programmatically.

   • [claimed-docs] “Get data that's displayed on the dashboard graphs in JSON format through the Dashboard REST API.”
   • [claimed-docs] “Create a dashboard”
   • [claimed-docs] “List all your cohorts in Amplitude, export a cohort in Amplitude, or upload a cohort.”
   • [claimed-docs] “Create, get, update, and delete categories, event types, event properties, and user properties.”

   proof ↗⚑ flag

2. product-managerCompose saved insights into shareable dashboards for my team

   weight 2 · round to PostHog
   PostHog✓fullclaimed8/10

   Docs explicitly state users can build insights, save them to dashboards, share them, and set alerts (posthog-docs-3, posthog-docs-25), directly matching the story of composing saved insights into shareable dashboards. Missing for 10: no independent/hands-on corroboration of the dashboard-sharing workflow specifically, and no detail on team-level permissions or collaboration features around shared dashboards.

   • [claimed-docs] “Build insights, assemble dashboards, share them, and set alerts on the numbers that matter.”
   • [claimed-docs] “Build trends, funnels, retention, paths, stickiness, and lifecycle insights on the events you already send to PostHog, then save them to das…”
   • [claimed-docs] “Build trends, funnels, retention, paths, stickiness, and lifecycle insights on the events you already send to PostHog”

   proof ↗⚑ flag

   Amplitude✓fullclaimed7/10

   Docs confirm core dashboard-building capabilities: creating dashboards, charts/insights, funnels/segments, and organizing dashboards/notebooks into team-owned spaces for collaboration, which directly supports composing insights into shareable team dashboards. Missing for 10: explicit documentation of sharing/permissions mechanics (e.g., link sharing, access controls) and independent/hands-on confirmation of the dashboard-building workflow beyond first-party docs.

   • [claimed-docs] “Charts, dashboards, and insights. Build funnels, segment users, answer "why".”
   • [claimed-docs] “Create a dashboard”
   • [claimed-docs] “Use spaces to organize charts, dashboards, and notebooks for the team that owns the work.”
   • [claimed-docs] “Build funnels, segment users, answer "why".”
   • [claimed-docs] “Get data that's displayed on the dashboard graphs in JSON format through the Dashboard REST API.”

   proof ↗⚑ flag

3. product-managerSubscribe to dashboards and get alerted when a key metric moves abnormally

   weight 1 · round to PostHog
   PostHog✓fullclaimed7/10

   PostHog docs explicitly state dashboards/insights can be saved, shared, and users can 'get alerted when they move' (posthog-docs-3, posthog-docs-25), and alerts can be routed to Slack/PagerDuty (posthog-intdir-2), directly matching the PM story of dashboard subscriptions with abnormal-metric alerts. Missing for 10: no independent/hands-on confirmation of the alerting UX, no detail on anomaly-detection thresholds, and no explicit mention of subscribing to an entire dashboard vs individual insights.

   • [claimed-docs] “Build insights, assemble dashboards, share them, and set alerts on the numbers that matter.”
   • [claimed-docs] “Build trends, funnels, retention, paths, stickiness, and lifecycle insights on the events you already send to PostHog, then save them to das…”
   • [claimed-docs] “Send alerts to Slack, PagerDuty, or monitoring tools”

   proof ↗⚑ flag

   Amplitude—none0/10

   The evidence pack covers dashboards, charts, cohorts, and a Dashboard REST API, but contains no mention of subscribing to dashboards or receiving alerts when a metric moves abnormally (e.g., anomaly detection, alerting rules, or subscription/notification features). missing for 10: evidence of dashboard subscription feature, evidence of anomaly/alert detection on metrics, evidence of notification delivery (email/Slack) tied to metric changes.

   • [claimed-docs] “Create a dashboard”
   • [claimed-docs] “Use spaces to organize charts, dashboards, and notebooks for the team that owns the work.”
   • [claimed-docs] “Get data that's displayed on the dashboard graphs in JSON format through the Dashboard REST API.”

   proof ↗⚑ flag

Identity

1. developerBulk-import historical events from another analytics tool or a data export into the platform

   weight 1 · round to Amplitude
   PostHog~partialcommunity3/10

   There's no first-party documentation of a dedicated bulk-import tool for historical events from another analytics tool, but the data warehouse/CDP features allow syncing external data sources into PostHog for unified querying, and a community comment notes Mixpanel-library compatibility making switching 'really easy.' missing for 10: explicit import API/tool for historical event data, documented data-export ingestion pipeline, and first-party confirmation of the Mixpanel-compatibility claim.

   • [claimed-docs] “Sync Stripe, Postgres, Salesforce, HubSpot, and dozens of other sources, then query them alongside your events, persons, and sessions in one…”
   • [claimed-docs] “It pulls data in from the tools you already use, lets you filter and reshape events as they're ingested, and sends the result wherever it ne…”
   • [community] “Great idea to be compatible with Mixpanel libraries! Makes switching over really easy”

   proof ↗⚑ flag

   Amplitude~partialclaimed4/10

   Amplitude's HTTP V2 API for sending events from a server (amplitude-docs-2) and third-party ETL/integration connectors like Fivetran and RudderStack (amplitude-intdir-5, amplitude-intdir-12) suggest a path for importing external or historical event data, but there is no explicit documentation of a dedicated bulk/batch historical import tool or backfill process. Missing for 10: explicit bulk-import/backfill API documentation, guidance on importing from competitor tools (e.g., Mixpanel/Segment exports), and any hands-on confirmation of successful historical data migration.

   • [claimed-docs] “Send data directly from your server to the HTTP V2 endpoint.”
   • [claimed-docs] “Fivetran is the leader in automated data integration, delivering ready-to-use connectors that adapt to change.”
   • [claimed-docs] “RudderStack provides Event streaming, ETL Pipelines, and Reverse ETL Pipelines that allow customers to have a 360 degree actionable view of …”
   • [claimed-docs] “Export your project's event data.”

   proof ↗⚑ flag

2. developerIdentify users across devices and merge anonymous pre-signup activity into their identified profile

   weight 2 · round to Amplitude
   PostHog—none0/10

   The evidence pack covers event capture, analytics, session replay, feature flags, experiments, and CDP integrations, but nowhere mentions identify()/alias() calls, distinct_id merging, or cross-device/anonymous-to-identified user reconciliation. This is a standard axis for an event-ingestion/analytics product, but no evidence supports it here. Missing for 10: any mention of identify/alias APIs, anonymous-to-known user merging, or cross-device identity resolution documentation.

   ⚑ flag

   Amplitude~partialclaimed3/10

   The only relevant evidence is a generic identity-model claim ('Every SDK shares the same identity, event, and consent model, so a user instrumented once is understood everywhere'), which gestures at cross-context identity resolution but never explicitly describes cross-device merging or converting anonymous pre-signup activity into an identified user profile. missing for 10: explicit documentation of user/device ID merge behavior, identify API or alias mechanism for anonymous-to-known user transition, and any independent confirmation of cross-device identity stitching.

   • [claimed-docs] “Every SDK shares the same identity, event, and consent model, so a user instrumented once is understood everywhere.”

   proof ↗⚑ flag

Instrumentation

1. product-managerCapture clicks, pageviews, and form interactions automatically without writing tracking code for each event

   weight 2 · round to Amplitude
   PostHog—none0/10

   The evidence pack discusses capturing custom events via SDKs/HTTP requests and building insights, but nowhere states that PostHog automatically captures clicks, pageviews, or form interactions without instrumenting code (no mention of autocapture). Missing for 10: any documented autocapture feature, evidence of automatic click/pageview/form tracking, or an example of a PM enabling capture with zero code.

   • [claimed-docs] “Custom events you capture from any SDK, each with the properties you attach to them.”
   • [claimed-docs] “capture events from any language that can send HTTP requests”

   proof ↗⚑ flag

   Amplitude~partialcommunity6/10

   Amplitude's docs claim built-in autocapture ('Auto-instrument page views, clicks, and sessions out of the box' and a Wizard CLI that 'instruments your codebase' in one command), directly matching the story of code-free automatic event capture. However, all support is first-party marketing/docs with no independent hands-on confirmation of the autocapture specifically, and community feedback (amplitude-comm-3) reports that real-world integration is 'very hard' and data is 'often wrong,' introducing real-world doubt about the ease claimed. Missing for 10: independent/hands-on verification that autocapture works as advertised without extra config, and resolution of the community-reported integration difficulty.

   • [claimed-docs] “Auto-instrument page views, clicks, and sessions out of the box.”
   • [claimed-docs] “Run the Wizard CLI to instrument your codebase and get a first dashboard from one terminal command.”
   • [claimed-docs] “The wizard detects your framework, installs the right SDK, adds your API key, and fires a test event.”
   • [community] “Amplitude is NOT a great product when you actually try to implement it. I have seen at least 3 companies pay for them... only to realize tha…”

   proof ↗⚑ flag

2. developerSend events from web, mobile, and backend apps through official SDKs for the major languages and platforms

   weight 3 · round to Amplitude
   PostHog✓fullcommunity7/10

   Docs confirm event capture from "any SDK" and from any HTTP-capable language, and community evidence corroborates real-world use of the iOS SDK integrating smoothly, supporting broad SDK coverage across platforms. Missing for 10: an explicit enumerated list of official SDKs per language/platform (web, Android, Node, Python, etc.) and independent confirmation for backend/server SDKs beyond the HTTP API and iOS mention.

   • [claimed-docs] “Custom events you capture from any SDK, each with the properties you attach to them.”
   • [claimed-docs] “capture events from any language that can send HTTP requests”
   • [community] “I've been using PostHog with my app for about a week now, and so far the results have been good. Pretty straightforward to integrate with a …”
   • [community] “Great idea to be compatible with Mixpanel libraries! Makes switching over really easy”

   proof ↗⚑ flag

   Amplitude✓fullclaimed8/10

   Docs explicitly describe a unified SDK model spanting web, mobile, server, and game engines with a shared identity/event model (amplitude-docs-36, amplitude-docs-16, amplitude-docs-37), plus a direct HTTP V2 endpoint for backend/server-side event sending (amplitude-docs-2) and auto-instrumentation for web/mobile (amplitude-docs-4). This covers the core of the story across major platforms via official SDKs. Missing for 10: an explicit enumerated list of supported languages/platforms and independent (non-vendor) confirmation of SDK reliability/breadth beyond docs claims.

   • [claimed-docs] “Send data directly from your server to the HTTP V2 endpoint.”
   • [claimed-docs] “Auto-instrument page views, clicks, and sessions out of the box.”
   • [claimed-docs] “Every SDK shares the same identity, event, and consent model, so a user instrumented once is understood everywhere.”
   • [claimed-docs] “Use Experiment SDKs to flag and test features across web, mobile, and backend.”
   • [claimed-docs] “Use the same API surface across web, mobile, server, and game engines.”
   • [claimed-docs] “Unified SDKs provide one integration point for multiple Amplitude products.”

   proof ↗⚑ flag

Ai analytics ops

1. ai-native userHave an agent create, target, and toggle feature flags through the API or MCP server

   weight 2 · round drawn
   PostHog~partialprobed6/10

   PostHog has an official MCP server (posthog-probe-4) and docs explicitly state that 'with just plain text questions your agents can ship a feature flag from a prompt' (posthog-docs-19, posthog-docs-24, posthog-docs-31), showing agent-driven flag creation via MCP. An OpenAPI spec exists (posthog-probe-3) implying REST API access to flags, but no evidence explicitly documents targeting rules or toggling flags via API/MCP endpoints specifically. missing for 10: explicit documentation of flag targeting (cohorts/properties) and toggle actions via MCP/API, and independent hands-on confirmation that agent-driven flag creation works reliably.

   • [claimed-docs] “with just plain text questions your agents can ship a feature flag from a prompt, dig into a stack trace without leaving your editor, run a …”
   • [claimed-docs] “your agents can ship a feature flag from a prompt, dig into a stack trace without leaving your editor, run a HogQL query through Claude”
   • [claimed-docs] “with just plain text questions your agents can ship a feature flag from a prompt, dig into a stack trace without leaving your editor”
   • [probe] “official MCP server documented at https://posthog.com/docs/model-context-protocol”
   • [probe] “PROBE openapi: HTTP 200 at https://posthog.com/openapi.json — contains "openapi" key”

   proof ↗⚑ flag

   Amplitude~partialclaimed6/10

   Amplitude's Experiment APIs let scripts/CI/tools create, target, and toggle flags (amplitude-docs-13, amplitude-docs-19, amplitude-docs-28, amplitude-docs-42), so an agent could drive flag lifecycle via API. However, the documented MCP server (amplitude-docs-11/17/34) is scoped to bringing analytics/behavioral context into AI tools for Q&A, not to flag creation/toggling, so the 'MCP server' half of the story is unsupported. Missing for 10: MCP-server tools for flag CRUD/targeting/toggling, explicit agent-oriented examples of calling the Experiment API, and any independent verification of this workflow.

   • [claimed-docs] “Manage flags, experiments, and deployments from CI, scripts, or your own tools with the Experiment APIs.”
   • [claimed-docs] “Release behind a flag, ramp to the audience you choose, and turn it off the moment something looks wrong.”
   • [claimed-docs] “Roll out behind a flag and run A/B tests so the data, not a hunch, drives the release.”
   • [claimed-docs] “Define the audience to target users by properties, cohorts, or segments.”
   • [claimed-docs] “Bring Amplitude’s behavioral context directly into your AI tools—Claude, Cursor, OpenAI, Lovable, Figma, and more—to get answers where you w…”
   • [claimed-docs] “Start a conversation about your users or product and get expert-level answers in seconds that you can share with your whole org. No SQL, con…”

   proof ↗⚑ flag

Experimentation

1. product-managerRun A/B experiments with goal metrics and statistical significance reported on the results

   weight 3 · round drawn
   PostHog✓fullclaimed8/10

   PostHog's Experiments docs explicitly describe defining variants, picking goal metrics, randomizing users, tracking exposures, and running statistics (Bayesian or frequentist) to determine significance, which directly matches the story. Missing for 10: independent/hands-on validation of the statistical reporting accuracy and no community corroboration specific to experiments (only analytics/session-replay community feedback exists).

   • [claimed-docs] “PostHog randomizes your users, tracks their exposures, and runs the statistics – Bayesian or frequentist – so you get a real answer instead …”
   • [claimed-docs] “You define variants, pick the metrics you care about, and PostHog randomizes your users, tracks their exposures, and runs the statistics – B…”
   • [claimed-docs] “You define variants, pick the metrics you care about, and PostHog randomizes your users, tracks their exposures, and runs the statistics – B…”
   • [claimed-docs] “PostHog randomizes your users, tracks their exposures, and runs the statistics – Bayesian or frequentist – so you get a real answer instead …”

   proof ↗⚑ flag

   Amplitude✓fullclaimed8/10

   Amplitude's Feature Experiment product explicitly supports A/B, multivariate, and bandit testing with audience targeting, causal lift measurement, and visual editing to ship variants, plus Experiment APIs for programmatic control (docs-25, docs-41, docs-42, docs-13, docs-15, docs-28). Statistical significance/causal lift is referenced ('measure causal lift') but no deep documentation of the specific stats methodology or independent hands-on validation of experiment result reporting is present. Missing for 10: detailed documentation of significance testing methodology/confidence intervals, and independent/hands-on corroboration of the experiment results UI.

   • [claimed-docs] “Ship with confidence. Test variants, target segments, measure causal lift.”
   • [claimed-docs] “Run A/B, multivariate, and bandit tests so the data, not the loudest opinion, makes the call.”
   • [claimed-docs] “Define the audience to target users by properties, cohorts, or segments.”
   • [claimed-docs] “Manage flags, experiments, and deployments from CI, scripts, or your own tools with the Experiment APIs.”
   • [claimed-docs] “Edit pages with a visual editor and ship the winning variant without a code deploy.”
   • [claimed-docs] “Roll out behind a flag and run A/B tests so the data, not a hunch, drives the release.”
   • [claimed-docs] “Use Experiment SDKs to flag and test features across web, mobile, and backend.”
   • [claimed-docs] “Release behind a flag, ramp to the audience you choose, and turn it off the moment something looks wrong.”

   proof ↗⚑ flag

Feature flags

1. developerRoll out features behind flags targeted by user properties, cohorts, and percentage rollouts

   weight 3 · round drawn
   PostHog✓fullclaimed8/10

   PostHog docs explicitly describe flagging changes and rolling out to a percentage of users, and targeting by person property, cohort, or group, which directly matches the story. missing for 10: no independent/hands-on corroboration of targeting accuracy or percentage rollout behavior beyond vendor docs.

   • [claimed-docs] “Wrap a change in a flag, roll it out to 1% of users, watch what happens, and turn it off the moment something looks wrong – no redeploy, no …”
   • [claimed-docs] “Roll out by person property, cohort, or group, then see the session replays, events, and exceptions from the people who got the flag”

   proof ↗⚑ flag

   Amplitude✓fullclaimed8/10

   Amplitude's Experiment/Feature Flags docs explicitly cover targeting by user properties, cohorts, and segments (docs-42), percentage-based ramping/rollouts (docs-19, docs-28), local evaluation for low latency (docs-3), and SDKs across web/mobile/server (docs-35), directly matching the story. missing for 10: independent/hands-on verification of flag targeting behavior beyond vendor docs, and no community evidence specifically confirming the flagging/rollout feature works as described.

   • [claimed-docs] “Evaluate flags locally for low-latency rollout decisions.”
   • [claimed-docs] “Release behind a flag, ramp to the audience you choose, and turn it off the moment something looks wrong.”
   • [claimed-docs] “Roll out behind a flag and run A/B tests so the data, not a hunch, drives the release.”
   • [claimed-docs] “Use Experiment SDKs to flag and test features across web, mobile, and backend.”
   • [claimed-docs] “Define the audience to target users by properties, cohorts, or segments.”
   • [claimed-docs] “Manage flags, experiments, and deployments from CI, scripts, or your own tools with the Experiment APIs.”

   proof ↗⚑ flag

Behavioral analysis

1. product-managerBuild multi-step funnels with filters and breakdowns to find where users drop off

   weight 3 · round to PostHog
   PostHog✓fullclaimed7/10

   PostHog docs explicitly state funnels are a first-class insight type built on captured events ('Build trends, funnels, retention, paths, stickiness...') alongside dashboards, alerts, and saved insights, which covers the core funnel-building use case for a PM. However, the evidence pack doesn't explicitly detail funnel-specific filter/breakdown UI mechanics beyond the general mention. Missing for 10: explicit documentation or screenshots of step-level filters and breakdown-by-property configuration within funnels, and independent user validation of funnel usability.

   • [claimed-docs] “Build trends, funnels, retention, paths, stickiness, and lifecycle insights on the events you already send to PostHog”
   • [claimed-docs] “Build trends, funnels, retention, paths, stickiness, and lifecycle insights on the events you already send to PostHog, then save them to das…”
   • [claimed-docs] “Custom events you capture from any SDK, each with the properties you attach to them.”
   • [claimed-docs] “Build insights, assemble dashboards, share them, and set alerts on the numbers that matter.”

   proof ↗⚑ flag

   Amplitude~partialcommunity7/10

   Docs explicitly reference building funnels, segmenting users, and answering 'why' via charts/dashboards, which is core to this story, but there is no direct documentation detail on breakdown/filter mechanics within a funnel, nor independent hands-on validation of funnel accuracy — one community comment even flags data accuracy concerns during implementation. missing for 10: explicit funnel filter/breakdown feature documentation, independent hands-on corroboration of funnel drop-off analysis working as described.

   • [claimed-docs] “Build funnels, segment users, answer "why".”
   • [claimed-docs] “Charts, dashboards, and insights. Build funnels, segment users, answer "why".”
   • [claimed-docs] “Create user cohorts”
   • [community] “Amplitude is NOT a great product when you actually try to implement it. I have seen at least 3 companies pay for them... only to realize tha…”

   proof ↗⚑ flag

2. product-managerAnalyze behavior at the account or company level, not just per user, for B2B products

   weight 2 · round to PostHog
   PostHog~partialclaimed4/10

   PostHog documents a group concept for targeting (person property, cohort, or group) used in feature flag rollouts, implying group-level segmentation exists, but the evidence pack never explicitly ties 'groups' to trends/funnels/retention insights or shows account-level rollups in analytics dashboards. missing for 10: explicit docs on group-level funnels/retention insights, account/company entity definition and configuration, examples of B2B account analytics dashboards.

   • [claimed-docs] “Roll out by person property, cohort, or group, then see the session replays, events, and exceptions from the people who got the flag”
   • [claimed-docs] “Build trends, funnels, retention, paths, stickiness, and lifecycle insights on the events you already send to PostHog”
   • [claimed-docs] “Build trends, funnels, retention, paths, stickiness, and lifecycle insights on the events you already send to PostHog, then save them to das…”

   proof ↗⚑ flag

   Amplitude—none0/10

   The evidence pack covers user-level analytics, cohorts, funnels, session replay, and experimentation, but contains no mention of account-level or company/group-based analytics (e.g., grouping events by an 'account' or 'company' entity) that B2B PMs need. Since account-based analysis is a plausible and common analytics-platform capability, its absence here is 'none' rather than 'na'.

   • [claimed-docs] “Build funnels, segment users, answer "why".”
   • [claimed-docs] “Create user cohorts”
   • [claimed-docs] “Charts, dashboards, and insights. Build funnels, segment users, answer "why".”
   • [claimed-docs] “Define the audience to target users by properties, cohorts, or segments.”

   proof ↗⚑ flag

3. product-managerMeasure retention over time and slice it by behavioral cohorts

   weight 3 · round to Amplitude
   PostHog✓fullclaimed7/10

   PostHog explicitly ships retention as a core insight type (posthog-docs-15, posthog-docs-25) alongside cohort support elsewhere in the product (posthog-docs-6 references rolling out and analyzing by cohort, posthog-docs-8/17 SQL for custom cohort queries). This directly matches the PM story of measuring retention and slicing by behavioral cohorts, though the evidence doesn't show a dedicated doc excerpt showing retention insight breakdown-by-cohort UI directly. missing for 10: an explicit doc/screenshot showing retention insight with cohort breakdown control, and independent hands-on confirmation of this specific workflow.

   • [claimed-docs] “Build trends, funnels, retention, paths, stickiness, and lifecycle insights on the events you already send to PostHog”
   • [claimed-docs] “Build trends, funnels, retention, paths, stickiness, and lifecycle insights on the events you already send to PostHog, then save them to das…”
   • [claimed-docs] “Roll out by person property, cohort, or group, then see the session replays, events, and exceptions from the people who got the flag”
   • [claimed-docs] “You can use SQL (Structured Query Language) throughout PostHog to manage, query, and modify data.”
   • [claimed-docs] “SQL insights enable you to directly query your data with SQL commands like SELECT, FROM, JOIN, WHERE, GROUP BY”

   proof ↗⚑ flag

   Amplitude✓fullclaimed8/10

   Amplitude is core analytics product with explicit retention/behavioral-cohort features: funnels/segmentation ('Build funnels, segment users, answer why'), user cohort creation, and cohort management APIs (list/export/upload). This directly supports retention-over-time analysis sliced by behavioral cohorts, which is a first-class Amplitude use case. Missing for 10: no explicit named 'Retention chart' documentation excerpt and no independent/hands-on report specifically confirming retention-by-cohort workflow quality.

   • [claimed-docs] “Build funnels, segment users, answer "why".”
   • [claimed-docs] “Create user cohorts”
   • [claimed-docs] “Charts, dashboards, and insights. Build funnels, segment users, answer "why".”
   • [claimed-docs] “List all your cohorts in Amplitude, export a cohort in Amplitude, or upload a cohort.”
   • [claimed-docs] “Create a dashboard”

   proof ↗⚑ flag

4. product-managerExplore the paths users actually take before or after a key event

   weight 2 · round to PostHog
   PostHog✓fullclaimed7/10

   PostHog explicitly documents 'paths' insights alongside trends, funnels, retention, stickiness, and lifecycle built on captured events (posthog-docs-15, posthog-docs-25), directly matching the path-analysis story. Missing for 10: no dedicated docs excerpt detailing path-specific configuration (e.g., start/end event anchoring, wildcard grouping) or independent/hands-on validation of the paths feature specifically.

   • [claimed-docs] “Build trends, funnels, retention, paths, stickiness, and lifecycle insights on the events you already send to PostHog”
   • [claimed-docs] “Build trends, funnels, retention, paths, stickiness, and lifecycle insights on the events you already send to PostHog, then save them to das…”
   • [claimed-docs] “Custom events you capture from any SDK, each with the properties you attach to them.”

   proof ↗⚑ flag

   Amplitude~partialclaimed5/10

   Amplitude's docs confirm funnel building and segmentation ('Build funnels, segment users, answer why') and session replay lets PMs see exact user sessions tied to funnel drop-off points, which supports exploring behavior around key events. However, no explicit mention of a dedicated Path/Journey analysis feature (e.g., Amplitude's Pathfinder/User Paths chart type) appears in the evidence, so the story's specific 'paths before/after an event' capability is only indirectly covered via funnels + session replay. Missing for 10: explicit path/journey analysis chart documentation, independent user validation of path exploration accuracy.

   • [claimed-docs] “Build funnels, segment users, answer "why".”
   • [claimed-docs] “Charts, dashboards, and insights. Build funnels, segment users, answer "why".”
   • [claimed-docs] “Go from a chart, funnel, or support ticket to the exact session.”
   • [claimed-docs] “Watch real user sessions to understand the why behind your metrics. Spot frustration, tie errors to UX, and measure engagement across every …”
   • [claimed-docs] “See exactly what users experience. Replay sessions with full event and network context.”

   proof ↗⚑ flag

Self serve insights

1. founderAnswer "which activation step loses users?" myself with funnels and drop-off analysis, without an analyst

   weight 3 · round drawn
   PostHog✓fullclaimed8/10

   PostHog's product analytics explicitly builds funnels, retention, trends, paths, stickiness, and lifecycle insights on captured events, with dashboards and alerts, all self-serve without requiring an analyst (posthog-docs-15, posthog-docs-25, posthog-docs-1). This directly supports self-service drop-off/activation analysis via funnels. missing for 10: no independent/hands-on evidence of a founder successfully using funnels to diagnose activation drop-off, and no case study demonstrating non-analyst usability in practice.

   • [claimed-docs] “Build trends, funnels, retention, paths, stickiness, and lifecycle insights on the events you already send to PostHog”
   • [claimed-docs] “Build trends, funnels, retention, paths, stickiness, and lifecycle insights on the events you already send to PostHog, then save them to das…”
   • [claimed-docs] “Custom events you capture from any SDK, each with the properties you attach to them.”
   • [claimed-docs] “Build insights, assemble dashboards, share them, and set alerts on the numbers that matter.”

   proof ↗⚑ flag

   Amplitude✓fullcommunity8/10

   Amplitude's core product is explicitly built around funnels, segmentation, and self-serve answers ('Build funnels, segment users, answer why') plus an AI assistant that answers plain-language questions without SQL or a data team, directly matching the founder self-serve need. Community evidence corroborates ease of adoption across non-analyst roles, though one review notes implementation/data-quality difficulties. Missing for 10: independent hands-on validation specifically of funnel/drop-off analysis accuracy and ease for a non-analyst founder persona.

   • [claimed-docs] “Build funnels, segment users, answer "why".”
   • [claimed-docs] “Charts, dashboards, and insights. Build funnels, segment users, answer "why".”
   • [claimed-docs] “Ask questions in plain language — Amplitude AI surfaces answers across all your data.”
   • [claimed-docs] “Start a conversation about your users or product and get expert-level answers in seconds that you can share with your whole org. No SQL, con…”
   • [community] “We got on for the generous free plan, stayed for the simple UX. Amplitude has been the only analytics product that we've gotten the whole te…”
   • [community] “Amplitude is NOT a great product when you actually try to implement it. I have seen at least 3 companies pay for them... only to realize tha…”

   proof ↗⚑ flag

Destinations

1. developerStream events onward to third-party destinations (CRM, ad platforms, webhooks) from the analytics platform

   weight 2 · round to PostHog
   PostHog✓fullclaimed8/10

   PostHog's CDP explicitly supports streaming events to third-party destinations including CRMs (Salesforce, HubSpot), ad platforms, webhooks, and SaaS tools (Braze, Customer.io, Slack, PagerDuty), with real-time or scheduled delivery via configurable Hog functions. missing for 10: independent/hands-on confirmation of destination reliability or throughput, and a full list of supported destinations beyond the named examples.

   • [claimed-docs] “It pulls data in from the tools you already use, lets you filter and reshape events as they're ingested, and sends the result wherever it ne…”
   • [claimed-docs] “It pulls data in from the tools you already use, lets you filter and reshape events as they're ingested, and sends the result wherever it ne…”
   • [claimed-docs] “Pipelines are built from Hog functions, which you can configure from a template, write yourself, or have PostHog AI write for you.”
   • [claimed-docs] “Sync user properties to CRM systems like Salesforce or HubSpot”
   • [claimed-docs] “Slack: Send events to Slack channels SaaS tools: Send events to SaaS tools like Braze, Customer.io, and more”
   • [claimed-docs] “Send alerts to Slack, PagerDuty, or monitoring tools”

   proof ↗⚑ flag

   Amplitude!disputedcontradicted6/10

   Amplitude's integrations directory documents connections to CRM (Salesforce, HubSpot), reverse-ETL/streaming tools (Hightouch, RudderStack, Segment), and states it can integrate with any system, supporting onward event streaming to third-party destinations. However, a hands-on community report explicitly states 'they dont do a good job in sending data back to things like SFDC,' directly contradicting the claimed CRM-sync capability for at least one major destination. Missing for 10: independent corroboration of reliable webhook/ad-platform streaming, and resolution of the documented SFDC sync complaint.

   • [claimed-docs] “Connect the leading data and marketing platforms to Amplitude. Don’t see what you need? We can integrate with any system.”
   • [claimed-docs] “Hightouch syncs data from data warehouses into tools that teams rely on to run the business (CRM, email, ERP, etc.”
   • [claimed-docs] “Salesforce provides customer relationship management software and applications focused on sales, customer service, marketing automation, ana…”
   • [claimed-docs] “RudderStack provides Event streaming, ETL Pipelines, and Reverse ETL Pipelines that allow customers to have a 360 degree actionable view of …”
   • [community] “Amplitude is NOT a great product when you actually try to implement it. I have seen at least 3 companies pay for them... only to realize tha…”

   proof ↗⚑ flag

1. ai-native userDo everything through the API that I can do in the UI

   weight 2 · round to PostHog
   PostHog✓fullprobed7/10

   PostHog exposes a full OpenAPI-documented REST API (posthog-probe-3, posthog-docs-11), an MCP server enabling agents to run queries, ship flags, and inspect errors (posthog-docs-2, -12, -19, -24, probe-4), a CLI for terminal/CI use (posthog-docs-20, probe-5), and SQL access throughout the product (posthog-docs-8, -17), collectively covering analytics, flags, experiments, error tracking, and data warehouse operations programmatically. missing for 10: an explicit vendor statement guaranteeing 1:1 API/UI parity, and independent verification that every UI workflow (e.g. survey building, dashboard sharing) is fully replicable via API without gaps.

   • [claimed-docs] “Run trends, funnels, retention, and SQL queries from any MCP client or AI editor.”
   • [claimed-docs] “capture events from any language that can send HTTP requests”
   • [claimed-docs] “"Show me the full stack trace for the most recent crash, then propose a fix." – Pulls the stack trace, error message, and metadata so the ag…”
   • [claimed-docs] “with just plain text questions your agents can ship a feature flag from a prompt, dig into a stack trace without leaving your editor, run a …”
   • [claimed-docs] “The PostHog CLI lets you use PostHog from your terminal, your coding agents, local scripts, and CI/CD pipelines.”
   • [claimed-docs] “You can use SQL (Structured Query Language) throughout PostHog to manage, query, and modify data.”
   • [probe] “PROBE openapi: HTTP 200 at https://posthog.com/openapi.json — contains "openapi" key”
   • [probe] “official MCP server documented at https://posthog.com/docs/model-context-protocol”
   • [probe] “official CLI documented at https://posthog.com/docs/cli”

   proof ↗⚑ flag

   Amplitude~partialprobed6/10

   Amplitude exposes broad APIs (HTTP V2 event ingestion, Dashboard REST API, Experiment APIs for flags/experiments, Cohorts API, taxonomy CRUD, data export) that mirror many core UI workflows, and an official MCP server lets AI tools query behavioral data. However, several flagship UI capabilities—session replay viewing/analysis, the visual A/B test editor, in-product guides/surveys creation, dashboard/chart creation, and the AI 'ask in plain language' analyst—are documented only as UI features with no corresponding API endpoints in the evidence pack. Missing for 10: API parity for session replay, visual experiment editor, dashboard/notebook creation, and guides/surveys authoring; independent confirmation that API coverage matches UI feature-for-feature.

   • [claimed-docs] “Send data directly from your server to the HTTP V2 endpoint.”
   • [claimed-docs] “Export your project's event data.”
   • [claimed-docs] “Manage flags, experiments, and deployments from CI, scripts, or your own tools with the Experiment APIs.”
   • [claimed-docs] “List all your cohorts in Amplitude, export a cohort in Amplitude, or upload a cohort.”
   • [claimed-docs] “Get data that's displayed on the dashboard graphs in JSON format through the Dashboard REST API.”
   • [claimed-docs] “Create, get, update, and delete categories, event types, event properties, and user properties.”
   • [claimed-docs] “Watch real user sessions to understand the why behind your metrics.”
   • [claimed-docs] “Edit pages with a visual editor and ship the winning variant without a code deploy.”
   • [claimed-docs] “Create a dashboard”
   • [probe] “PROBE openapi: HTTP 200 at https://amplitude.com/openapi.json — contains "openapi" key”
   • [probe] “official MCP server documented at https://amplitude.com/mcp-server”

   proof ↗⚑ flag

2. ai-native userExport all of my data in open formats and leave

   weight 3 · round to PostHog
   PostHog~partialprobed6/10

   PostHog is open-source (MIT-licensed self-host Docker Compose) and exposes data via SQL, an HTTP API, and data-warehouse sync, which together give strong data portability and self-hosting options implying users can retain full control of their data. However, there is no explicit documented 'export all your data' feature or open-format bulk export tool, and no independent evidence of someone successfully exporting/migrating a full production dataset. Missing for 10: explicit bulk-export/download-all-data documentation, confirmation of open standard export formats (e.g., CSV/Parquet dumps), and independent hands-on evidence of a full data migration/exit.

   • [claimed-docs] “PostHog is open-source and freely available for anyone to host themselves. We offer a free Docker Compose deployment under an MIT license.”
   • [claimed-docs] “We offer a free Docker Compose deployment under an MIT license.”
   • [claimed-docs] “You can use SQL (Structured Query Language) throughout PostHog to manage, query, and modify data.”
   • [claimed-docs] “SQL insights enable you to directly query your data with SQL commands like SELECT, FROM, JOIN, WHERE, GROUP BY”
   • [claimed-docs] “Sync Stripe, Postgres, Salesforce, HubSpot, and dozens of other sources, then query them alongside your events, persons, and sessions in one…”
   • [claimed-docs] “capture events from any language that can send HTTP requests”
   • [probe] “PROBE openapi: HTTP 200 at https://posthog.com/openapi.json — contains "openapi" key”

   proof ↗⚑ flag

   Amplitude~partialclaimed5/10

   Amplitude provides data export mechanisms (event data export, Dashboard REST API returning JSON, cohort export/upload) via documented APIs, which support portability of raw data. However, there's no evidence of a comprehensive full-account export (e.g., all configuration, dashboards, experiments, session replays) in a single open-format bundle, nor any documented account-deletion/data-portability workflow for 'leaving' the platform. missing for 10: full-account export tooling covering dashboards/configs/experiments/session-replay data, an explicit data-portability/account-closure workflow, and independent verification that exported formats are truly open and complete.

   • [claimed-docs] “Export your project's event data.”
   • [claimed-docs] “List all your cohorts in Amplitude, export a cohort in Amplitude, or upload a cohort.”
   • [claimed-docs] “Get data that's displayed on the dashboard graphs in JSON format through the Dashboard REST API.”

   proof ↗⚑ flag

3. ai-native userRead the product's source under an open license

   weight 2 · round to PostHog
   PostHog✓fullclaimed9/10

   PostHog explicitly documents its open-source nature and MIT-licensed self-host deployment (Docker Compose), with the actual source hosted publicly on GitHub, satisfying the ability to read the product's source under an open license. Missing for 10: no explicit confirmation of the license covering the entire codebase (vs just the self-host component) or independent audit of license terms.

   • [claimed-docs] “We offer a free Docker Compose deployment under an MIT license.”
   • [claimed-docs] “PostHog is open-source and freely available for anyone to host themselves. We offer a free Docker Compose deployment under an MIT license.”

   proof ↗⚑ flag

   Amplitude—none0/10

   Amplitude is a closed-source, proprietary SaaS analytics platform; the evidence pack shows only some SDKs (e.g., Flutter) are open-source but no evidence the core product's source is open-licensed or readable. This axis applies to any product (a buyer could ask about source availability), but there's no evidence of an open license covering Amplitude's core codebase.

   • [claimed-docs] “The Flutter SDK is an open-source UI for developing applications for Android, iOS, Windows, Mac, Linux, Google Fuchsia and the Web.”

   proof ↗⚑ flag

4. ai-native userSelf-host the core product

   weight 3 · round to PostHog
   PostHog!disputedcontradicted5/10

   PostHog's docs explicitly state the core product is open-source and offers a free MIT-licensed Docker Compose self-host deployment (posthog-docs-14/21), which is a clear ai-native/openness claim. However, a hands-on community report describes the production Dockerfile/posthog:latest image as 'busted' and failing to create a database, requiring significant debugging effort — a concrete contradiction of the self-host claim working out of the box. Missing for 10: independent confirmation the current self-host path works reliably, resolution or acknowledgment of the reported Docker issue, and details on feature parity between self-hosted and cloud versions.

   • [claimed-docs] “We offer a free Docker Compose deployment under an MIT license.”
   • [claimed-docs] “PostHog is open-source and freely available for anyone to host themselves. We offer a free Docker Compose deployment under an MIT license.”
   • [community] “You should mention on the README that the production dockerfile (and posthog/posthog:latest) are busted, they do not create any database. Sp…”

   proof ↗⚑ flag

   Amplitude—none0/10

   Amplitude is offered exclusively as a hosted SaaS analytics platform; no evidence of a self-hostable core product, open-source release, or on-prem deployment option appears anywhere in the docs, integrations, or community threads.

   ⚑ flag

Privacy first analytics

1. founderTrack product usage without third-party cookies so I can minimize or avoid cookie consent banners

   weight 2 · round to Amplitude
   PostHog—none0/10

   The evidence pack contains no mention of cookie-based vs cookieless tracking, consent banners, or privacy-compliant capture configuration—only general analytics, session replay, and SDK/API capture details. Without any documentation addressing cookieless tracking or consent-banner avoidance, this axis has no supporting evidence.

   ⚑ flag

   Amplitude~partialclaimed3/10

   Amplitude's docs mention server-side event sending (HTTP V2 endpoint) and a shared consent model with options to mask sensitive content and honor consent, which implies some capacity for privacy-conscious, non-cookie-dependent tracking, but there is no explicit claim about avoiding third-party cookies or reducing cookie-consent banners. missing for 10: explicit documentation of cookieless/first-party tracking, cookie policy details, or guidance on reducing consent-banner requirements.

   • [claimed-docs] “Send data directly from your server to the HTTP V2 endpoint.”
   • [claimed-docs] “Every SDK shares the same identity, event, and consent model, so a user instrumented once is understood everywhere.”
   • [claimed-docs] “Mask sensitive content, honor consent, and control what gets captured.”

   proof ↗⚑ flag

1. ai-native userChoose where my data is stored (region/residency)

   weight 2 · round drawn
   PostHog—none0/10

   The evidence pack has self-hosting/open-source docs (posthog-docs-14, posthog-docs-21) which relate to deployment flexibility but no mention of data residency/region selection (e.g., EU vs US cloud region choice) for PostHog Cloud. Missing for 10: explicit region/residency selection options, data storage location controls, compliance documentation on where data resides.

   • [claimed-docs] “We offer a free Docker Compose deployment under an MIT license.”
   • [claimed-docs] “PostHog is open-source and freely available for anyone to host themselves. We offer a free Docker Compose deployment under an MIT license.”

   proof ↗⚑ flag

   Amplitude—none0/10

   No evidence in the pack mentions data residency, region selection, EU/US hosting options, or any data storage location controls for Amplitude. This is a fair axis for an analytics SaaS platform handling user data, but absence of evidence means it cannot be credited as delivered. missing for 10: any mention of regional data hosting options, EU/US residency controls, or data location settings.

   ⚑ flag

2. ai-native userPrevent my data from being used to train AI models

   weight 3 · round drawn
   PostHog—none0/10

   No evidence in the pack addresses AI-training data opt-out or any privacy control preventing customer data from being used to train AI models; nothing in the docs or community items mentions this capability.

   ⚑ flag

   Amplitude—none0/10

   No evidence in the pack addresses AI-training opt-out or data-usage-for-AI-training controls; the pack covers analytics, session replay, feature flags, and an MCP server but nothing about preventing data from being used to train AI models.

   ⚑ flag

3. ai-native userControl data retention and deletion

   weight 2 · round to Amplitude
   PostHog—none0/10

   The evidence pack contains no mention of data retention settings, deletion APIs, GDPR-style data deletion, or retention policy controls anywhere in the docs, CLI, API, or CDP descriptions. This is a fair and expected axis for an analytics/data platform handling user event data, so absence of evidence means the story is unmet rather than inapplicable.

   ⚑ flag

   Amplitude~partialclaimed3/10

   Docs mention consent controls and content masking for session replay (amplitude-docs-33) and CRUD including delete for event/user properties via API (amplitude-docs-38), which touch on privacy controls, but there's no explicit documentation of a user-level data retention policy or GDPR-style deletion/export mechanism for end-user data. missing for 10: explicit data retention policy settings, a documented user data deletion/erasure API or workflow, and any independent confirmation that deletion requests are honored.

   • [claimed-docs] “Mask sensitive content, honor consent, and control what gets captured.”
   • [claimed-docs] “Create, get, update, and delete categories, event types, event properties, and user properties.”

   proof ↗⚑ flag

4. ai-native userOpt out of telemetry and usage tracking

   weight 2 · round to Amplitude
   PostHog—none0/10

   The evidence pack covers PostHog's product analytics, session replay, feature flags, MCP/CLI integrations, and self-hosting, but contains no mention of an opt-out setting for PostHog's own telemetry/usage tracking (e.g., CLI or SDK phone-home data collection). No documentation or community evidence addresses this capability.

   ⚑ flag

   Amplitude~partialclaimed4/10

   Amplitude documents a shared 'consent model' across SDKs and features to 'honor consent, and control what gets captured' for session replay, implying some opt-out/consent tooling exists in its data-collection stack, but there is no dedicated documentation about an AI-native user opting out of Amplitude's own telemetry or usage tracking of the product itself. missing for 10: explicit opt-out/DNT mechanism documentation, privacy policy specifics for AI-native/developer users, independent confirmation that consent controls actually block tracking.

   • [claimed-docs] “Every SDK shares the same identity, event, and consent model, so a user instrumented once is understood everywhere.”
   • [claimed-docs] “Mask sensitive content, honor consent, and control what gets captured.”

   proof ↗⚑ flag

Replay

1. product-managerJump from a funnel drop-off or an error event straight to session replays of the affected users

   weight 2 · round to Amplitude
   PostHog~partialclaimed6/10

   PostHog documents linking session replays directly to flag rollouts and to exceptions/errors ('see the session replays, events, and exceptions from the people who got the flag'), and session replay's synced DevTools panel shows errors at the moment they happened, supporting a PM-style jump from error/event to replay. However, the evidence never explicitly documents jumping from a funnel drop-off step to session replays, only from feature-flag cohorts/exceptions. missing for 10: explicit funnel-to-replay linking documentation, independent/hands-on confirmation of this specific workflow.

   • [claimed-docs] “Roll out by person property, cohort, or group, then see the session replays, events, and exceptions from the people who got the flag”
   • [claimed-docs] “Session replay records what real users do in your product and plays it back like a DVR, with a synced DevTools panel showing console logs, n…”
   • [claimed-docs] “with a synced DevTools panel showing console logs, network requests, and errors at the exact moment they happened”
   • [claimed-docs] “plays it back like a DVR, with a synced DevTools panel showing console logs, network requests, and errors at the exact moment they happened”

   proof ↗⚑ flag

   Amplitude✓fullclaimed9/10

   Amplitude's session replay docs explicitly describe going from a chart, funnel, or support ticket to the exact session, catching errors and rage clicks across sessions, and tying errors to UX — directly matching the funnel-drop-off/error-to-replay jump story. This is well documented across multiple docs pages with clear first-party detail on the workflow. missing for 10: independent/hands-on third-party corroboration that the funnel-to-replay jump works smoothly in practice.

   • [claimed-docs] “Go from a chart, funnel, or support ticket to the exact session.”
   • [claimed-docs] “Catch rage clicks, dead clicks, and errors across sessions.”
   • [claimed-docs] “Watch real user sessions to understand the why behind your metrics. Spot frustration, tie errors to UX, and measure engagement across every …”
   • [claimed-docs] “See exactly what users experience. Replay sessions with full event and network context.”
   • [claimed-docs] “Focus replays on specific users, pages, and flows.”

   proof ↗⚑ flag

2. product-managerWatch recordings of real user sessions with sensitive input masked by default

   weight 3 · round to Amplitude
   PostHog—none0/10

   Evidence confirms PostHog's session replay lets PMs watch DVR-style recordings with synced DevTools panel (posthog-docs-26, posthog-docs-29, posthog-gh-8), but nothing in the pack mentions input masking, privacy controls, or sensitive-data redaction being applied by default. missing for 10: any documentation of input masking/privacy defaults, evidence of what is masked or how, and confirmation it's on by default rather than opt-in.

   • [claimed-docs] “Session replay records what real users do in your product and plays it back like a DVR, with a synced DevTools panel showing console logs, n…”
   • [claimed-docs] “plays it back like a DVR, with a synced DevTools panel showing console logs, network requests, and errors at the exact moment they happened”
   • [github] “Session replays: Watch real user sessions of interactions with your website or mobile app to diagnose issues and understand user behavior.”

   proof ↗⚑ flag

   Amplitude✓fullclaimed8/10

   Amplitude's session replay docs explicitly state you can watch real user sessions ([amplitude-docs-8],[amplitude-docs-21],[amplitude-docs-24]) and that sensitive content is masked by default with consent controls ([amplitude-docs-33]), directly matching the story. Missing for 10: independent/hands-on verification of default masking behavior and more detail on configurability/exceptions.

   • [claimed-docs] “Watch real user sessions to understand the why behind your metrics.”
   • [claimed-docs] “Watch real user sessions to understand the why behind your metrics. Spot frustration, tie errors to UX, and measure engagement across every …”
   • [claimed-docs] “See exactly what users experience. Replay sessions with full event and network context.”
   • [claimed-docs] “Mask sensitive content, honor consent, and control what gets captured.”
   • [claimed-docs] “Go from a chart, funnel, or support ticket to the exact session.”
   • [claimed-docs] “Focus replays on specific users, pages, and flows.”

   proof ↗⚑ flag

Ai analytics ops

1. ai-native userHave an agent answer growth questions from live product data by running queries through the API or MCP server

   weight 3 · round to Amplitude
   PostHog!disputedcontradicted5/10

   PostHog documents a first-party MCP server and CLI that let agents run HogQL/SQL queries, ship flags, and answer product questions from live data (posthog-docs-2, posthog-docs-19, posthog-docs-24, posthog-probe-4, posthog-docs-8/17 for SQL, posthog-docs-9 for warehouse queries). However, a hands-on community report describes the LLM/query integration failing to create a cohort via a SQL query and causing severe UI slowdowns, directly contradicting the smooth 'agent answers growth questions via query' promise (posthog-comm-8). missing for 10: independent corroboration that MCP-driven SQL queries reliably work at scale, and resolution of the reported failure/performance bug.

   • [claimed-docs] “Run trends, funnels, retention, and SQL queries from any MCP client or AI editor.”
   • [claimed-docs] “with just plain text questions your agents can ship a feature flag from a prompt, dig into a stack trace without leaving your editor, run a …”
   • [claimed-docs] “your agents can ship a feature flag from a prompt, dig into a stack trace without leaving your editor, run a HogQL query through Claude”
   • [probe] “official MCP server documented at https://posthog.com/docs/model-context-protocol”
   • [claimed-docs] “Sync Stripe, Postgres, Salesforce, HubSpot, and dozens of other sources, then query them alongside your events, persons, and sessions in one…”
   • [claimed-docs] “SQL insights enable you to directly query your data with SQL commands like SELECT, FROM, JOIN, WHERE, GROUP BY”
   • [community] “I really wish you guys would do something about the llm integration, I tried using it two days ago to create a cohort of users using a sql q…”

   proof ↗⚑ flag

   Amplitude✓fullprobed8/10

   Amplitude ships an official MCP server that lets AI tools (Claude, Cursor, OpenAI, etc.) query behavioral data conversationally without SQL or dashboards, plus REST/Dashboard/Export APIs for programmatic access to live product data. Together these enable an agent to answer growth questions from live data via API or MCP. Missing for 10: independent hands-on verification of the MCP server's query accuracy/reliability and no community corroboration of agent-driven analysis in practice.

   • [claimed-docs] “Bring Amplitude’s behavioral context directly into your AI tools—Claude, Cursor, OpenAI, Lovable, Figma, and more—to get answers where you w…”
   • [claimed-docs] “Bring Amplitude's behavioral context directly into your AI tools—Claude, Cursor, OpenAI, Lovable, Figma, and more—to get answers where you w…”
   • [claimed-docs] “Start a conversation about your users or product and get expert-level answers in seconds that you can share with your whole org. No SQL, con…”
   • [claimed-docs] “Get data that's displayed on the dashboard graphs in JSON format through the Dashboard REST API.”
   • [claimed-docs] “Create, get, update, and delete categories, event types, event properties, and user properties.”
   • [probe] “official MCP server documented at https://amplitude.com/mcp-server”
   • [probe] “PROBE openapi: HTTP 200 at https://amplitude.com/openapi.json — contains "openapi" key”

   proof ↗⚑ flag

Raw access

1. developerRun analytical queries and pull raw event data back through a documented query/export API

   weight 3 · round drawn
   PostHog✓fullprobed8/10

   PostHog documents a full HogQL SQL query interface, a data warehouse that lets you query events/persons alongside external sources, a public OpenAPI spec (confirmed live at /openapi.json), and API-based event capture/export, plus MCP support for running queries programmatically. This directly satisfies running analytical queries and pulling raw event data via a documented API. missing for 10: independent/hands-on confirmation of raw event export via REST API specifically (evidence is mostly first-party docs), and no third-party validation of API rate limits/completeness for large-scale export.

   • [claimed-docs] “You can use SQL (Structured Query Language) throughout PostHog to manage, query, and modify data.”
   • [claimed-docs] “SQL insights enable you to directly query your data with SQL commands like SELECT, FROM, JOIN, WHERE, GROUP BY”
   • [claimed-docs] “Sync Stripe, Postgres, Salesforce, HubSpot, and dozens of other sources, then query them alongside your events, persons, and sessions in one…”
   • [claimed-docs] “capture events from any language that can send HTTP requests”
   • [probe] “PROBE openapi: HTTP 200 at https://posthog.com/openapi.json — contains "openapi" key”
   • [claimed-docs] “your agents can ship a feature flag from a prompt, dig into a stack trace without leaving your editor, run a HogQL query through Claude”

   proof ↗⚑ flag

   Amplitude✓fullprobed8/10

   Amplitude documents a Dashboard REST API returning JSON query results, an Export API for raw event data, and other CRUD APIs (events, cohorts, taxonomy), with an OpenAPI spec confirmed reachable — covering both analytical queries and raw event export as the story requires. Missing for 10: independent/hands-on developer corroboration of the export/query API's reliability and rate limits beyond first-party docs.

   • [claimed-docs] “Send data directly from your server to the HTTP V2 endpoint.”
   • [claimed-docs] “Export your project's event data.”
   • [claimed-docs] “Get data that's displayed on the dashboard graphs in JSON format through the Dashboard REST API.”
   • [claimed-docs] “Create, get, update, and delete categories, event types, event properties, and user properties.”
   • [probe] “PROBE openapi: HTTP 200 at https://amplitude.com/openapi.json — contains "openapi" key”

   proof ↗⚑ flag

2. developerQuery my raw events with SQL (or an SQL-equivalent query language) inside the platform

   weight 3 · round to PostHog
   PostHog✓fullclaimed9/10

   PostHog documents HogQL, a SQL-like query language, that lets developers directly query raw events with SELECT/FROM/JOIN/WHERE/GROUP BY, and also join across data warehouse sources alongside events/persons/sessions in one SQL query, plus MCP/agent access to run HogQL queries. This is well-documented first-party capability with clear syntax examples. missing for 10: independent hands-on verification of SQL query correctness/performance at scale (one community comment notes LLM confusion with cohorts, not the SQL feature itself).

   • [claimed-docs] “You can use SQL (Structured Query Language) throughout PostHog to manage, query, and modify data.”
   • [claimed-docs] “SQL insights enable you to directly query your data with SQL commands like SELECT, FROM, JOIN, WHERE, GROUP BY”
   • [claimed-docs] “Sync Stripe, Postgres, Salesforce, HubSpot, and dozens of other sources, then query them alongside your events, persons, and sessions in one…”
   • [claimed-docs] “with just plain text questions your agents can ship a feature flag from a prompt, dig into a stack trace without leaving your editor, run a …”
   • [claimed-docs] “your agents can ship a feature flag from a prompt, dig into a stack trace without leaving your editor, run a HogQL query through Claude”

   proof ↗⚑ flag

   Amplitude—none0/10

   Evidence shows Amplitude offers dashboards, charts, cohorts, and an AI natural-language query interface (which explicitly markets 'No SQL' as a benefit), plus REST/export APIs and warehouse integrations like BigQuery/Snowflake for data export — but nothing indicates an in-platform SQL or SQL-equivalent query editor over raw events. missing for 10: any documented SQL/notebook query interface inside Amplitude, evidence of a query language for ad-hoc raw event access, or hands-on confirmation of SQL support.

   • [claimed-docs] “Ask questions in plain language — Amplitude AI surfaces answers across all your data.”
   • [claimed-docs] “Start a conversation about your users or product and get expert-level answers in seconds that you can share with your whole org. No SQL, con…”
   • [claimed-docs] “Export your project's event data.”
   • [claimed-docs] “Get data that's displayed on the dashboard graphs in JSON format through the Dashboard REST API.”
   • [claimed-docs] “BigQuery is Google's fully managed, petabyte scale, low cost analytics data warehouse.”
   • [claimed-docs] “Snowflake makes it easy to amass all your data, enable rapid analytics, and derive data-driven insights for all your business users.”

   proof ↗⚑ flag

Warehouse

1. developerSync events between the platform and my own data warehouse (query warehouse tables in-product or export events continuously to it)

   weight 2 · round to PostHog
   PostHog✓fullclaimed8/10

   PostHog's data warehouse feature lets developers sync external sources (Stripe, Postgres, Salesforce, HubSpot, etc.) and query them alongside native events/persons/sessions in one SQL query, and the CDP allows continuous export of events to external destinations in realtime or on a schedule, satisfying both directions of the sync story. Missing for 10: independent/hands-on verification of the warehouse sync reliability or performance at scale, and clearer documentation of specific export destinations (data warehouses) beyond CRM/Slack examples.

   • [claimed-docs] “Sync Stripe, Postgres, Salesforce, HubSpot, and dozens of other sources, then query them alongside your events, persons, and sessions in one…”
   • [claimed-docs] “You can use SQL (Structured Query Language) throughout PostHog to manage, query, and modify data.”
   • [claimed-docs] “SQL insights enable you to directly query your data with SQL commands like SELECT, FROM, JOIN, WHERE, GROUP BY”
   • [claimed-docs] “It pulls data in from the tools you already use, lets you filter and reshape events as they're ingested, and sends the result wherever it ne…”
   • [claimed-docs] “Pipelines are built from Hog functions, which you can configure from a template, write yourself, or have PostHog AI write for you.”
   • [claimed-docs] “Combine everything in PostHog's context warehouse so that you, your agents, and your dashboard can query it directly.”

   proof ↗⚑ flag

   Amplitude~partialclaimed6/10

   Amplitude documents an Export API for event data and lists first-party integrations with warehouses (BigQuery, Snowflake, Databricks) plus reverse-ETL tools (Hightouch, Fivetran, RudderStack) that support syncing events to/from a warehouse, satisfying the 'continuous export' half of the story. However, there's no evidence of an in-product warehouse query feature (e.g., querying warehouse tables directly within Amplitude's UI). Missing for 10: in-product warehouse querying capability, and independent/hands-on confirmation of the export-to-warehouse pipelines actually working reliably.

   • [claimed-docs] “Export your project's event data.”
   • [claimed-docs] “BigQuery is Google's fully managed, petabyte scale, low cost analytics data warehouse.”
   • [claimed-docs] “Databricks is a unified, open analytics platform that enables building, deploying, sharing, and maintaining enterprise-grade data, analytics…”
   • [claimed-docs] “Snowflake makes it easy to amass all your data, enable rapid analytics, and derive data-driven insights for all your business users.”
   • [claimed-docs] “Hightouch syncs data from data warehouses into tools that teams rely on to run the business (CRM, email, ERP, etc.”
   • [claimed-docs] “Fivetran is the leader in automated data integration, delivering ready-to-use connectors that adapt to change.”

   proof ↗⚑ flag