Skip to content
MCP server

ChainMemory

Portable project memory for AI agents, anchored with verifiable hashes

Runs locallyStatus and history
Pricing
Freemium
Platforms
CLI, API, Chrome extension
Runs
Locally
Version
2.10.0
Listing updated
2 days ago

About ChainMemory

ChainMemory is a freemium MCP server for knowledge and memory. It runs locally on your machine and is started by your AI client.

Memory is organised as a Project Brain: decisions, milestones, risks, priorities, constraints and open questions for each project. Rather than loading everything, get_project_brief builds plain text within a character budget (default 7,000), listing what changed since the last read, open risks, active priorities and recent decisions. A task mode narrows the brief to one priority. Briefs are safe to paste into chats, since IP addresses, ports, server paths and security rules are replaced with [withheld].

Version 2.10.0 adds an optional blind vault. With a 12-word BIP-39 phrase set locally, sealed memories are encrypted on your machine and their search vector is computed there, so searches send a vector instead of text. The phrase never leaves the machine and cannot be recovered. A full-detail brief needs an explicit environment flag and the owner's key, and each such request is logged.

An API key is required for every tool except sealing. The service also offers a Chrome extension and a REST API, and the code is MIT licensed.

Install

Run in a terminal

claude mcp add chainmemory --env CHAINMEMORY_API_KEY='<CHAINMEMORY_API_KEY>' --env CHAINMEMORY_SEED_PHRASE='<CHAINMEMORY_SEED_PHRASE>' --env AICHAIN_KEY='<AICHAIN_KEY>' -- npx -y chainmemory-mcp

Add to claude_desktop_config.json

{
  "mcpServers": {
    "chainmemory": {
      "command": "npx",
      "args": [
        "-y",
        "chainmemory-mcp"
      ],
      "env": {
        "CHAINMEMORY_API_KEY": "<CHAINMEMORY_API_KEY>",
        "CHAINMEMORY_SEED_PHRASE": "<CHAINMEMORY_SEED_PHRASE>",
        "AICHAIN_KEY": "<AICHAIN_KEY>"
      }
    }
  }
}

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "chainmemory": {
      "command": "npx",
      "args": [
        "-y",
        "chainmemory-mcp"
      ],
      "env": {
        "CHAINMEMORY_API_KEY": "<CHAINMEMORY_API_KEY>",
        "CHAINMEMORY_SEED_PHRASE": "<CHAINMEMORY_SEED_PHRASE>",
        "AICHAIN_KEY": "<AICHAIN_KEY>"
      }
    }
  }
}

Add to .vscode/mcp.json

{
  "servers": {
    "chainmemory": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "chainmemory-mcp"
      ],
      "env": {
        "CHAINMEMORY_API_KEY": "<CHAINMEMORY_API_KEY>",
        "CHAINMEMORY_SEED_PHRASE": "<CHAINMEMORY_SEED_PHRASE>",
        "AICHAIN_KEY": "<AICHAIN_KEY>"
      }
    }
  }
}

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

{
  "mcpServers": {
    "chainmemory": {
      "command": "npx",
      "args": [
        "-y",
        "chainmemory-mcp"
      ],
      "env": {
        "CHAINMEMORY_API_KEY": "<CHAINMEMORY_API_KEY>",
        "CHAINMEMORY_SEED_PHRASE": "<CHAINMEMORY_SEED_PHRASE>",
        "AICHAIN_KEY": "<AICHAIN_KEY>"
      }
    }
  }
}

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

{
  "mcpServers": {
    "chainmemory": {
      "command": "npx",
      "args": [
        "-y",
        "chainmemory-mcp"
      ],
      "env": {
        "CHAINMEMORY_API_KEY": "<CHAINMEMORY_API_KEY>",
        "CHAINMEMORY_SEED_PHRASE": "<CHAINMEMORY_SEED_PHRASE>",
        "AICHAIN_KEY": "<AICHAIN_KEY>"
      }
    }
  }
}

Configuration

  • CHAINMEMORY_API_KEYChainMemory API key. REQUIRED for every tool except chainmemory_seal. Get one at https://faucet.chainmemory.aiRequired
  • CHAINMEMORY_SEED_PHRASE12-word BIP-39 phrase for the blind vault. Optional: without it the server works exactly as before. With it, sealed:true encrypts a memory on your machine and computes its search vector here, so sealed memories can still be found, and searches send the query's vector instead of its text. Generate one with chainmemory_new_seed. It never leaves your machine and cannot be recovered.
  • CHAINMEMORY_API_BASEAPI base URL (default: https://api.chainmemory.ai)
  • CHAINMEMORY_MODELS_DIRFolder for the 45 MB search model used with the blind vault (default: ~/.chainmemory/models). Downloaded once from models.chainmemory.ai and checked against its SHA-256 anchored on ChainMemory's chain.
  • CHAINMEMORY_ALLOW_FULLSet to 1 to let get_project_brief return the FULL level (addresses, ports, server paths and security rules) when you ask for it. Leave it empty to keep every brief at the public level, which is safe to paste anywhere. Works only with the owner's key, and ChainMemory logs every full-level request.
  • AICHAIN_KEYNOT NEEDED for normal use — leave it empty. Only the chainmemory_seal tool reads it, to sign a contract call yourself instead of letting the API anchor the memory (which chainmemory_remember already does). If you do set it, it is a wallet private key: use a dedicated wallet holding only a few AIC, never your main one. No other tool uses it.
  • AICHAIN_RPCRPC endpoint used only by chainmemory_seal (default: https://rpc.chainmemory.ai). Leave it empty unless you set AICHAIN_KEY.

Tools this server provides (9)

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

get_project_brief
Plain-text project summary within a character budget
get_project_state
Full project state
update_project_state
Update project state with a new verifiable fingerprint
save_memory
Save a memory linked to a project
search_memories
Search memories by content, tags, project or date
inject_context
Inject relevant context into a conversation
verify_state_hash
Verify a past state against its hash
chainmemory_seal
Seal a memory
chainmemory_new_seed
Generate a 12-word seed phrase

What it does

Project Brain state
Tracks decisions, priorities, risks and milestones per project.
Budgeted briefs
Builds a prioritised summary within a set character limit.
Blind vault
Encrypts sealed memories locally with a 12-word phrase.
Verifiable fingerprints
Each state change produces a hash that can be verified later.
Cross-model memory
The same memories work with ChatGPT, Claude, Gemini and Perplexity.
Redacted briefs
Server paths, ports and addresses are withheld by default.

Good for

  • Resume a project in a new chat without re-explaining it
  • Move context between Claude and ChatGPT
  • Prove what a project's state was at an earlier point
  • Keep sensitive notes encrypted on your own machine

Pricing

Freemium

Free plan: 100 memories per month; paid tiers are Starter, Pro and Team, paid in USDT

Prices change. Check chainmemory.ai before you buy.

ChainMemory FAQ

Is an API key required?

Yes, for every tool except chainmemory_seal; keys come from the project's faucet site.

Can I use more than one model?

Yes. The Chrome extension covers ChatGPT, Claude, Gemini and Perplexity, and the MCP server works with MCP clients such as Claude Desktop and Claude Code.

How are plans priced?

Free, Starter, Pro and Team tiers differ by monthly memories, injects, projects and retention; prices appear in USD, paid today in USDT.

Can the seed phrase be recovered?

No. The README says it never leaves your machine and cannot be recovered.

Is ChainMemory free?

ChainMemory has a free plan, with paid plans for more usage or features.

How do I add ChainMemory to Claude Code?

Run this command in your terminal: claude mcp add chainmemory --env CHAINMEMORY_API_KEY='<CHAINMEMORY_API_KEY>' --env CHAINMEMORY_SEED_PHRASE='<CHAINMEMORY_SEED_PHRASE>' --env AICHAIN_KEY='<AICHAIN_KEY>' -- npx -y chainmemory-mcp

What are the best alternatives to ChainMemory?

The closest alternatives to ChainMemory are Codebase Memory, XERJ, Claude Code Ultimate Guide, Nautilus Compass Memory and note1.

Alternatives to ChainMemory

  1. 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
  2. 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
  3. 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
  4. 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
  5. note1

    MCP server

    MCP server giving AI assistants search and summaries across your note1 meetings

    • 4 tools
    • Remote
    • Free trial
    • 1 GitHub star
    • Productivity