🤖 Connect Agents via MCP

The bundled Model Context Protocol server lets any MCP-capable agent — Claude Code, Claude Desktop, or custom SDK agents — store and recall memory autonomously.

Claude Code

claude mcp add lighthouse-memory \
  --env LIGHTHOUSE_API_KEY=YOUR_API_KEY \
  -- npx -y @lighthouse-web3/memory

Add --scope user to make it available in every project. Pass additional env vars (MEMORY_NETWORK, MEMORY_NAMESPACE, …) with more --env flags.

Claude Desktop / other MCP clients

Add to the client's MCP configuration:

{
  "mcpServers": {
    "lighthouse-memory": {
      "command": "npx",
      "args": ["-y", "@lighthouse-web3/memory"],
      "env": { "LIGHTHOUSE_API_KEY": "YOUR_API_KEY" }
    }
  }
}

npx -y @lighthouse-web3/memory runs the package's bundled MCP server (lighthouse-memory-mcp) over stdio.

Available tools

Tool	What it does
memory_remember	Store content (+tags, metadata); buffers locally, flushes in batches
memory_flush	Force all pending memories onto the network as one batch blob
memory_recall	Semantic + keyword search, optional tag filter
memory_list	Newest-first listing
memory_get	Full record by id or CID
memory_forget	Remove from index; deletes the blob once no other memory shares it
memory_rebuild_index	Restore the index from the network on a new machine
memory_status	Counts (incl. pending), network, embeddings mode, snapshot CID

What it looks like in practice

Once connected, the agent uses memory naturally:

You: remember that our staging environment is at staging.acme.dev and redeploys on every merge to develop

Agent: (calls memory_remember) Saved.

— days later, new session —

You: where do I check the latest staging build?

Agent: (calls memory_recall) Your staging environment is at staging.acme.dev — it redeploys on every merge to develop.

tip

Encourage durable habits in your agent's system prompt, e.g. "Before ending a session, call memory_flush if any memories are pending."