MCP Server

Use OpenClaw Cortex as a memory provider in Claude Desktop or any MCP-compatible client

OpenClaw Cortex implements the Model Context Protocol (MCP), allowing it to be used as a memory provider directly from Claude Desktop or any MCP-compatible client.

Start the MCP server#

openclaw-cortex mcp

The server communicates over stdio (stdin/stdout), which is the standard MCP transport. It does not bind to a network port.

Configure in Claude Desktop#

Add the following to your Claude Desktop configuration file:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Linux: ~/.config/claude/claude_desktop_config.json

{
  "mcpServers": {
    "openclaw-cortex": {
      "command": "openclaw-cortex",
      "args": ["mcp"],
      "env": {
        "ANTHROPIC_API_KEY": "sk-ant-...",
        "OPENCLAW_CORTEX_MEMGRAPH_URI": "bolt://localhost:7687"
      }
    }
  }
}

After adding this configuration, restart Claude Desktop. You should see "openclaw-cortex" listed as an available MCP server.

Available Tools#

remember#

Store a memory in Memgraph.

| Parameter | Type | Required | Description | |---|---|---|---| | content | string | yes | The text content to remember | | type | string | no | Memory type: rule, fact, episode, procedure, or preference (default: fact) | | scope | string | no | Memory scope: permanent, project, session, or ttl (default: permanent) | | project | string | no | Project name for project-scoped memories | | confidence | number | no | Confidence score 0.0–1.0 (default: 1.0) |

Example interaction:

Claude will call remember with:

{
  "content": "Always use snake_case for database column names",
  "type": "rule",
  "scope": "project",
  "project": "my-project"
}

Response:

{
  "id": "550e8400-e29b-41d4-a716-446655440000",
  "stored": true
}

recall#

Retrieve memories relevant to a message using semantic search, graph traversal, and multi-factor ranking.

| Parameter | Type | Required | Description | |---|---|---|---| | message | string | yes | The query to recall memories for | | project | string | no | Project context for scope boosting | | budget | number | no | Token budget for returned context (default: 2000) |

Response:

{
  "context": "--- Relevant Memories ---\n[rule] Always use snake_case for database columns...\n",
  "memory_count": 2
}

forget#

Delete a memory by ID.

| Parameter | Type | Required | Description | |---|---|---|---| | id | string | yes | The UUID of the memory to delete |

Response:

{
  "deleted": true
}

search#

Semantic search over memories. Returns raw results with similarity scores, without multi-factor re-ranking.

| Parameter | Type | Required | Description | |---|---|---|---| | message | string | yes | The query to search for | | limit | number | no | Maximum number of results (default: 10) | | project | string | no | Filter results to this project |

Response:

{
  "results": [
    {
      "memory": {
        "id": "550e8400-...",
        "content": "Always use snake_case for database columns",
        "type": "rule",
        "scope": "project"
      },
      "score": 0.934
    }
  ]
}

stats#

Get statistics about the memory collection.

Response:

{
  "total_memories": 87,
  "by_type": {
    "rule": 14,
    "fact": 42,
    "episode": 20,
    "procedure": 8,
    "preference": 3
  },
  "by_scope": {
    "permanent": 55,
    "project": 18,
    "session": 10,
    "ttl": 4
  }
}

Usage Patterns#

Automatic context injection#

Ask Claude to recall memories before working on a task:

Saving important decisions#

Searching past work#

Requirements#

Memgraph and Ollama must be running before starting the MCP server. The remember tool does not require ANTHROPIC_API_KEY — only the CLI capture command uses the Anthropic API.

docker compose up -d           # start Memgraph
ollama pull nomic-embed-text   # ensure embedding model is available
openclaw-cortex mcp            # start MCP server

Troubleshooting#

"MCP server not appearing in Claude Desktop": Restart Claude Desktop after editing the config file. Check that openclaw-cortex is in your PATH.

"Error: connection refused": Ensure Memgraph is running (docker compose up -d) and Ollama is running (ollama serve). Memgraph listens on Bolt port 7687.

"Tool call failed": Check that OPENCLAW_CORTEX_MEMGRAPH_URI and OPENCLAW_CORTEX_OLLAMA_BASE_URL point to the right hosts if you are not running services locally.