Skip to content
MCP server

Heartwood Memory — governed memory for AI agents

Governed local memory for coding agents with signed provenance and policy-gated recall

Runs locallyStatus and history
Install
Pricing
Freemium
Platforms
CLI, self-hosted, Linux
Runs
Locally
Version
0.2.9
Listing updated
1 day ago

About Heartwood Memory — governed memory for AI agents

Heartwood Memory — governed memory for AI agents is a freemium MCP server for knowledge and memory. It runs locally on your machine and is started by your AI client.

Heartwood targets forgetting between sessions and projects rather than inside one context window. Every memory is signed and the signature is re-checked at each recall. Before ranking, recall removes records the caller is not cleared to use, and each result carries its author, source IDs and signature status. Writes and deletions go into a hash-chained audit log, and erasing a subject destroys that subject's key and can issue a signed receipt verifiable offline.

It runs as an embedded Python library or a local stdio MCP server over one SQLite file shared by clients such as Claude Code, Codex and a read-only Copilot plugin. By default the tool allowlist is read-only: recall, explain_recall and health, while remember and forget must be named explicitly. If no database path is set, memory lives in RAM and is lost when the server stops.

The core is source-available under BSL 1.1, not OSI open source. Community use is free for non-production and for organizations under 100 people and $1 million in revenue; Team is $349 a month and Professional $6,000 a year, offered to US organizations.

Install

Run in a terminal

claude mcp add heartwood-memory --env HEARTWOOD_DB_PATH='<HEARTWOOD_DB_PATH>' -- uvx heartwood-memory

Add to claude_desktop_config.json

{
  "mcpServers": {
    "heartwood-memory": {
      "command": "uvx",
      "args": [
        "heartwood-memory"
      ],
      "env": {
        "HEARTWOOD_DB_PATH": "<HEARTWOOD_DB_PATH>"
      }
    }
  }
}

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "heartwood-memory": {
      "command": "uvx",
      "args": [
        "heartwood-memory"
      ],
      "env": {
        "HEARTWOOD_DB_PATH": "<HEARTWOOD_DB_PATH>"
      }
    }
  }
}

Add to .vscode/mcp.json

{
  "servers": {
    "heartwood-memory": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "heartwood-memory"
      ],
      "env": {
        "HEARTWOOD_DB_PATH": "<HEARTWOOD_DB_PATH>"
      }
    }
  }
}

Add to ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "heartwood-memory": {
      "command": "uvx",
      "args": [
        "heartwood-memory"
      ],
      "env": {
        "HEARTWOOD_DB_PATH": "<HEARTWOOD_DB_PATH>"
      }
    }
  }
}

Add to your client's MCP settings (most use this shape)

{
  "mcpServers": {
    "heartwood-memory": {
      "command": "uvx",
      "args": [
        "heartwood-memory"
      ],
      "env": {
        "HEARTWOOD_DB_PATH": "<HEARTWOOD_DB_PATH>"
      }
    }
  }
}

Configuration

  • HEARTWOOD_DB_PATHPath to the local Heartwood SQLite database.Required
  • HEARTWOOD_TENANTTenant identifier used to isolate memories.
  • HEARTWOOD_MCP_ALLOWED_TOOLSComma-separated MCP tool allowlist; the default is read-only.

Tools this server provides (5)

Taken from the project documentation; the running server may differ.

recall
Recalls memories the caller is cleared to use.
explain_recall
Explains why a recall result was returned.
health
Reports server health.
remember
Stores a memory; must be added to the allowlist.
forget
Erases memories; must be added to the allowlist.

What it does

Signed memories
Signs each memory and rechecks the signature on every recall.
Policy-gated recall
Removes records the caller is not cleared for before ranking.
Crypto-shred erasure
Destroys a subject's key and can issue a signed receipt.
Hash-chained audit log
Records writes and deletions tamper-evidently.
Read-only by default
Exposes recall, explain_recall and health unless you add more.
Shared local store
One SQLite file readable by several MCP clients.
Opt-in hosted judge
Hosted plans can use an optional model to rank recall.

Good for

  • Give coding agents memory that carries across sessions and projects
  • Show where an agent's answer came from
  • Erase one person's data and keep proof of it
  • Share one memory store between Claude Code and Codex
  • Audit what an agent remembered and recalled

Pricing

Freemium, from $349 a month

Community free for non-production and small organizations; Team $349/month; Professional $6,000/year; Enterprise quoted

Prices change. Check heartwoodmemory.com before you buy.

Heartwood Memory — governed memory for AI agents FAQ

Is it open source?

No. The core is source-available under BSL 1.1, and each release becomes Apache-2.0 four years after shipping.

What does it cost?

Community is free for non-production and for qualifying small organizations; Team is $349/month and Professional is $6,000/year.

Does it keep memory after a restart?

Only if you set an absolute database path; otherwise memory stays in RAM.

Is hosted MCP available?

No. MCP runs over local stdio on one machine.

Is Heartwood Memory — governed memory for AI agents free?

Heartwood Memory — governed memory for AI agents has a free plan, with paid plans for more usage or features. Paid plans start at $349 a month.

How do I add Heartwood Memory — governed memory for AI agents to Claude Code?

Run this command in your terminal: claude mcp add heartwood-memory --env HEARTWOOD_DB_PATH='<HEARTWOOD_DB_PATH>' -- uvx heartwood-memory

What are the best alternatives to Heartwood Memory — governed memory for AI agents?

The closest alternatives to Heartwood Memory — governed memory for AI agents are REA, Codebase Memory, XERJ, Claude Code Ultimate Guide and Nautilus Compass Memory.

Alternatives to Heartwood Memory — governed memory for AI agents

  1. REA

    MCP server

    CLI and MCP server that lets coding agents reverse engineer apps and binaries

    Runs locally
    • Open source
    • 464K GitHub stars
    • Developer tools
  2. Codebase Memory

    MCP server

    Local MCP server that builds a code knowledge graph for AI coding agents

    Runs locally
    • Open source
    • 48.7K GitHub stars
    • Knowledge and memory
  3. XERJ

    MCP server

    Local search engine for AI agents with cited answers over your code and docs

    Runs locally
    • Open source
    • 19.2K GitHub stars
    • Knowledge and memory
  4. MCP server for searching the Claude Code Ultimate Guide from any MCP client

    Runs locally
    • Open source
    • 6.5K GitHub stars
    • AI models and LLM tooling
  5. Local-first long-term memory layer for AI agents, served over MCP

    Not checked yet
    • 5 tools
    • Remote
    • Open source
    • 1.3K GitHub stars
    • Knowledge and memory
  6. Agent Roast

    MCP server

    Offline review that flags agent changes which weaken tests, assertions or CI

    Runs locally
    • Open source
    • 4 Hacker News points
    • Developer tools