Memory for AI agents

One self-hosted service your agents search by meaning, over REST, MCP, OAuth or the CLI.

Try it in your browser GitHub PyPI
recallexample data
>
    5 ms
    local read latency
    86%
    LongMemEval R@5, session mode
    5,043
    tests
    25+
    AI apps supported
    Every memory is a node, every relationship an edge. Recall works by meaning, not keywords.
    See it in 50 seconds: store, retrieve, link and consolidate. Narrated with a local text-to-speech model.

    What's new in v12.3

    Visible Codex memory activity, reliable inbox delivery, and a bootstrap authorization fix.

    • Agents

      Reliable OpenCode inbox polling

      A configurable overlap window catches messages around polling boundaries. Previously delivered rows are dropped so each message is shown once.

    • Hooks

      Visible Codex memory activity

      Codex lifecycle hooks report memory load and save status through system messages, so successful hook activity is visible in the session.

    • Health

      Cloudflare sync failures are visible

      Health checks report when Cloudflare cannot be reached instead of leaving an unavailable sync service looking healthy.

    • Storage

      Milvus memory searches work again

      Milvus accepts the shared agent_id argument without failing on ordinary searches. Agent-scoped searches return an explicit unsupported-operation result.

    • Dashboard

      Consistent dashboard controls

      Toggles, sliders, and checkboxes use the dashboard's cyan accent consistently.

    • Security

      Bootstrap reads preserve session state

      The bootstrap tool now requires write scope. Read-only clients can use the bootstrap resource without resetting session counters or changing access metadata.

    Full release notes for v12.3.0

    One service, three ways in

    Agents on your machine talk to it over stdio, frameworks over REST, and browser chat apps over Remote MCP with OAuth 2.1.

    Terminal and IDE agentsClaude Code, Codex CLI, Cursor, OpenCode
    Local MCPstdio, on your machine
    Agent frameworksLangGraph, CrewAI, AutoGen, any HTTP client
    REST APIHTTP with API key or OAuth
    Browser chatclaude.ai, ChatGPT (Developer Mode)
    Remote MCPStreamable HTTP, OAuth 2.1 + PKCE
    Memory service Search by meaning, knowledge graph, consolidation SQLite-vec, Cloudflare or Hybrid storage

    The underlined lane is the one that crosses the network with OAuth. Local MCP is a process on your machine, with no network hop.

    The architecture diagram: stdio, REST and Streamable HTTP entries leading to MemoryServer and the storage backends Architecture diagramThe three ways in, with the source lines behind every box. The deployment diagram: GitHub, its Actions runner, PyPI, Docker Hub, Cloudflare Pages and the GitLab mirror Deployment landscapeWhere releases and the site are built, published and mirrored.

    In the wild

    Three deployments people built and run themselves.

    "After I work with one of the cluster agents on something I want my local agent to know about, the cluster agent adds a special tag to the memory entry that my local agent recognizes as a message from a cluster agent. So they end up using it as a comms bridge, and it's pretty delightful."
    @jeremykoerber

    A five-agent cluster uses one memory service as shared state and as an inter-agent channel, with no custom protocol: cluster agents write a sentinel tag, the local agent filters on it. The pattern

    "File-based memory demands constant discipline. Semantic recall from a live database doesn't. Storing data on my own hardware while making it remotely accessible across platforms turned out to be a feature I didn't know I needed."
    @PL-Peter

    Two Docker containers on one SQLite backend, a Cloudflare tunnel for name-based routing, and an auth gateway in front. The checklist

    "Claude Code, Claude Desktop, Codex CLI and OpenCode all talk to the same sqlite-vec DB over stdio on my Mac. ~5,900 memories and counting. It's the closest thing to persistent identity my agents have."
    Mingjian Shao, AI PM and consultant, September 2026

    One local sqlite-vec database shared by four agents: no server, no cloud, offline embeddings, daily backups and scheduled consolidation.

    How it compares

    Against the commercial memory APIs and the MCP-native alternatives.

    Commercial memory APIs

    mcp-memory-serviceMem0ZepDIY Redis+Pinecone
    LicenseApache 2.0ProprietaryEnterprise—
    Cost$0Per-call APIEnterpriseInfra costs
    claude.ai browserRemote MCPDesktop onlyDesktop onlyNo
    OAuth 2.0 + DCRYesUnknownUnknownNo
    Streamable HTTPYes (SSE too)NoNoNo
    Framework integrationREST, any HTTP clientSDKSDKManual
    Knowledge graphYes, typed edgesNoLimitedNo
    Auto consolidationYes, decay + compressionNoNoNo
    On-premise embeddingsYes, ONNX localNoNoManual
    Privacy100% localCloudCloudPartial
    Hybrid searchYes, BM25 + vectorNoYesManual
    MCP protocolYesNoNoNo

    MCP-native: MemPalace

    mcp-memory-serviceMemPalace
    LongMemEval R@5 (raw ChromaDB, zero LLM)86.0% session / 80.4% turn96.6%1
    LongMemEval R@5 (with reranking)—100%2
    Storage granularityTurn-level and session-levelSession-level
    Team / multi-device syncCloudflare syncLocal only
    REST API / web dashboardYesNo
    OAuth 2.1 + multi-userYesNo
    Knowledge graphYes, typed edgesNo
    Auto consolidationYes, decay + compressionNo
    LicenseApache 2.0MIT

    Why the benchmark gap? MemPalace stores whole sessions as single units, and LongMemEval asks which session contains the answer, so that granularity answers the question structurally. This service defaults to turn-level storage for finer retrieval; session-mode ingestion brings it to 86.0% R@5.
    1 Measured in MemPalace "raw mode": plain text in ChromaDB with default embeddings. Per Issue #27, the Palace structural features are bypassed in that configuration, and the maintainers acknowledged most of the review's points.
    2 Uses optional LLM reranking (~500 API calls) on a partially tuned test set. Clean held-out score as reported by the maintainers: 98.4% R@5.
    Full numbers and the run commands: docs/BENCHMARKS.md

    SHODH ecosystem

    Fully compatible with the SHODH Unified Memory API Specification: every SHODH implementation shares the same memory schema (emotional metadata, episodic memory, source tracking, quality scoring), so memories export and import across implementations with full fidelity. The reference implementation is shodh-memory (RocksDB); shodh-cloudflare runs on Workers + Vectorize; this service runs SQLite-vec or Hybrid.

    In the media

    Talks and coverage from the community.

    • Heavybit, Humans in the Loop
      Agents Overdrawn at the Memory Bank

      A deep dive with maintainer Heinrich Krupp on agent amnesia and why persistent memory is the missing infrastructure layer for agentic systems.

    • AI Tinkerers Zürich
      Talk by Heinrich Krupp · full video

      Persistent memory architecture, semantic search, and multi-agent memory sharing.

    • LinkedIn, Dr. Álvaro Cintas
      Solving session amnesia

      How mcp-memory-service gives Claude Code, Cursor and OpenCode a shared memory over MCP, with on-device ONNX embeddings and no external API calls.

    "Your project has inspired me in many ways. In my view, it's the best implementation of MCP memory I've found so far." — Michał Zubkowicz

    Get started in seconds

    Install from PyPI and connect your agents to persistent memory.

    GitHub Documentation Discord
    The live demo: a stored memory list and a search that returns ranked results, all in the browser Live demoStore and search memories in your browser. Nothing is uploaded. The tutorial: how to add persistent memory to claude.ai in five minutes Set up claude.ai in 5 minutesRemote MCP with OAuth, step by step.

    Deploying for a team?I help teams run self-hosted memory in production: OAuth, hybrid sync, multi-client rollout. Free 30-minute intro call.

    Get in touch