AI Memory Layers arenaBuyer checklist
Every requirement we judge ai memory layers products against, as a ready-to-send RFP checklist — with each item's priority, why it matters, and how the top-ranked products score on it today.
57 requirements · 13 themes · verdicts for 6 products · updated 2026-09-16 · priorities mirror the story weights our scoring uses (methodology)
Show the markdown export
# AI Memory Layers — buyer checklist (RFP) Derived from ProductArena's evidence-graded user-story taxonomy for AI Memory Layers: 57 judged requirements. Priorities mirror story weights (3 = must-have, 2 = should-have, 1 = nice-to-have). ## Agenticness - [ ] **[must-have]** Plug MCP servers into this product so it can use their tools - [ ] **[must-have]** Connect an agent via an official MCP server - [ ] **[must-have]** Drive the product through a documented public API - [ ] **[must-have]** Delegate tasks to a built-in AI assistant inside the product - [ ] **[should-have]** Point an agent at llms.txt or agent-oriented docs - [ ] **[should-have]** Run the product headlessly / in CI for automation - [ ] **[should-have]** Use an official CLI - [ ] **[should-have]** Issue scoped/least-privilege API credentials for an agent - [ ] **[should-have]** Build against official SDKs - [ ] **[should-have]** Subscribe to events via webhooks - [ ] **[should-have]** Get AI-generated insights and suggestions from my data inside the product - [ ] **[should-have]** Set up automations that run autonomously in the background - [ ] **[should-have]** Operate the product with natural-language commands - [ ] **[should-have]** Explore an interactive API reference with runnable examples - [ ] **[should-have]** Download a machine-readable API spec (OpenAPI or equivalent) - [ ] **[should-have]** Rely on versioned APIs with a documented deprecation policy - [ ] **[nice-to-have]** Test against a sandbox environment without touching production data ## Automation depth - [ ] **[must-have]** Define rules that trigger actions automatically on events - [ ] **[should-have]** Perform bulk operations across many items at once - [ ] **[should-have]** Schedule recurring jobs or workflows - [ ] **[nice-to-have]** Version, review, and roll back my automations ## Data lifecycle - [ ] **[should-have]** Delete a user's memories on demand — single memory, per-entity, or full erasure — to satisfy privacy requirements - [ ] **[should-have]** Make memories expire or decay — via TTL, expiration dates, or recency weighting — so stale facts stop surfacing - [ ] **[should-have]** Export memories in a machine-readable format so the memory store is portable and not a lock-in trap ## Deployment self host - [ ] **[must-have]** Self-host the memory layer from open-source code (e.g. via Docker) on infrastructure I control - [ ] **[nice-to-have]** Run the memory layer fully locally — embedded in-process or against local models — without any cloud dependency ## Graph entity memory - [ ] **[must-have]** Store memories as a knowledge graph of entities and relationships so multi-hop and entity-centric questions are answerable - [ ] **[should-have]** Track when facts became valid or invalid (temporal reasoning) so the memory distinguishes current from outdated information - [ ] **[nice-to-have]** Customize the memory schema — entity types, edge types, or ontology — to match my domain ## Memory recall quality - [ ] **[must-have]** My agent can manage its own memory mid-conversation — adding, searching, updating, and deleting memories through tools or API calls it invokes itself - [ ] **[must-have]** Add memories from conversations and retrieve them later with semantic search, so context persists across sessions - [ ] **[should-have]** The memory layer decides for itself what is worth remembering — extracting salient facts from raw conversation and consolidating them in the background - [ ] **[should-have]** See published memory-quality benchmark results (e.g. LongMemEval, LoCoMo) backing the product's recall-accuracy claims - [ ] **[should-have]** Rely on the memory layer to update, supersede, or merge memories when new information contradicts what was stored - [ ] **[should-have]** Steer retrieval with metadata filters, keyword/hybrid search modes, or reranking instead of accepting a single fixed similarity search ## Openness - [ ] **[must-have]** Export all of my data in open formats and leave - [ ] **[must-have]** Self-host the core product - [ ] **[should-have]** Do everything through the API that I can do in the UI - [ ] **[should-have]** Read the product's source under an open license ## Pricing plans - [ ] **[should-have]** See published pricing with a free tier and per-unit rates so I can project memory costs before committing ## Privacy posture - [ ] **[must-have]** Prevent my data from being used to train AI models - [ ] **[should-have]** Choose where my data is stored (region/residency) - [ ] **[should-have]** Control data retention and deletion - [ ] **[should-have]** Opt out of telemetry and usage tracking ## Retrieval performance - [ ] **[should-have]** See documented retrieval-latency targets or measured numbers (e.g. p50/p95) backing the product's speed claims - [ ] **[should-have]** Ingest at scale with async or batch processing and check the status of background memory operations ## Sdk integrations - [ ] **[should-have]** Connect off-the-shelf assistants (Claude, ChatGPT, Cursor) to the same memory so every tool I use shares what it knows about me - [ ] **[should-have]** Drop the memory layer into agent frameworks like LangChain, LangGraph, CrewAI, or the Vercel AI SDK via documented first-party integrations - [ ] **[should-have]** Build against official SDKs in at least Python and TypeScript with equivalent memory APIs - [ ] **[nice-to-have]** Wire memory into real-time voice pipelines (e.g. LiveKit, Pipecat, ElevenLabs) with documented integrations fast enough for live conversation ## Session context - [ ] **[must-have]** Get summaries of past sessions or threads so an agent can pick up where the last conversation left off - [ ] **[should-have]** Retrieve a token-budgeted, prompt-ready context block assembled from relevant memories in one call - [ ] **[should-have]** Ingest documents, JSON, and business data into memory — not just chat transcripts - [ ] **[nice-to-have]** Store images, PDFs, or other files as memory inputs and recall information from them later ## Tenancy permissions - [ ] **[must-have]** Scope memories per user, agent, or application so one tenant's memories never leak into another's retrieval - [ ] **[should-have]** Govern who and what can read or write memory with roles, policies, or access-control lists, and audit that access - [ ] **[nice-to-have]** Share selected memory across multiple agents or users (team or group memory) while keeping private memory private --- Source: https://ultrametric.ai/productarena/arena/ai-memory (evidence-graded verdicts for 6 products) · methodology: https://ultrametric.ai/productarena/methodology
Chips show the top 5 ranked products' current verdict on each requirement — ✓ full · ~ partial · ! disputed · — none · n/a not applicable.
Agenticness — how well agents can access and operate the productAgenticness· 17 items
How well agents can access and operate the product
Automation depth — how much of the product can run unattendedAutomation depth· 4 items
How much of the product can run unattended
Data lifecycle — stories about data lifecycle in this arenaData lifecycle· 3 items
Stories about data lifecycle in this arena
Deployment self host — stories about deployment self host in this arenaDeployment self host· 2 items
Stories about deployment self host in this arena
Graph entity memory — stories about graph entity memory in this arenaGraph entity memory· 3 items
Stories about graph entity memory in this arena
Memory recall quality — stories about memory recall quality in this arenaMemory recall quality· 6 items
Stories about memory recall quality in this arena
- should-have
ai-native userThe memory layer decides for itself what is worth remembering — extracting salient facts from raw conversation and consolidating them in the background
Important, not disqualifying — weighs 2× in arena scoring · 2 of 5 products fully deliver this today
Openness — open source, data portability, and self-hosting storiesOpenness· 4 items
Open source, data portability, and self-hosting stories
Pricing plans — plan structure and value — what each tier costs and what it unlocksPricing plans· 1 item
Plan structure and value — what each tier costs and what it unlocks
Privacy posture — data-handling and privacy storiesPrivacy posture· 4 items
Data-handling and privacy stories
Retrieval performance — stories about retrieval performance in this arenaRetrieval performance· 2 items
Stories about retrieval performance in this arena
Sdk integrations — stories about sdk integrations in this arenaSdk integrations· 4 items
Stories about sdk integrations in this arena
Session context — stories about session context in this arenaSession context· 4 items
Stories about session context in this arena
Tenancy permissions — stories about tenancy permissions in this arenaTenancy permissions· 3 items
Stories about tenancy permissions in this arena
Full evidence behind every verdict lives on the arena page and each product page — chips above deep-link straight to the judged story.