Deployment

Deployment Options

twenty-mcp supports local stdio, a loopback-only development listener, a least-privilege stdio container, and a remote authenticated HTTPS boundary. Provider endpoint, credential, selector, identity, and trust material are supplied at runtime through AgentConfig; none is stored in this repository.

Installed stdio process

{
  "mcpServers": {
    "twenty": {
      "command": "twenty-mcp",
      "args": [],
      "env": {"MCP_TOOL_MODE": "intent"}
    }
  }
}

Loopback development listener

twenty-mcp --transport streamable-http --host 127.0.0.1 --port 8000

Do not expose this listener beyond loopback. Network deployments require direct TLS or an explicitly trusted TLS-terminating ingress, configured authentication, exact MCP_ALLOWED_HOSTS, and an exact trusted-proxy CIDR policy.

Least-privilege local container

docker run -i --rm \
  --read-only \
  --cap-drop=ALL \
  --security-opt=no-new-privileges \
  --pids-limit=256 \
  --tmpfs /tmp:rw,noexec,nosuid,nodev,size=64m \
  -e TRANSPORT=stdio \
  registry.example.invalid/twenty-mcp@sha256:<digest> twenty-mcp

The operator projects the selected AgentConfig profile into the process at runtime; the image remains immutable and contains no environment connection profile.

Remote authenticated HTTPS endpoint

{
  "mcpServers": {
    "twenty": {"url": "https://service.example.invalid/mcp"}
  }
}

Store the real remote URL, outbound identity reference, and TLS-profile reference in AgentConfig, not in MCP client JSON or documentation.

This page covers running twenty-mcp as a long-lived service: the transports, the optional A2A agent server, a Docker Compose stack, putting it behind a Caddy reverse proxy, and giving it a DNS name with Technitium. To provision the Twenty CRM platform it connects to, see Backing Platform.

twenty-mcp ships an MCP server (console script twenty-mcp) and an A2A agent server (console script twenty-agent) that delegates CRM work to the MCP tools. Deploy the MCP server alone, or both together.

Run the MCP server

The transport is selected with --transport (or the TRANSPORT env var):

stdio (default)streamable-httpsse

twenty-mcp

For IDE / desktop MCP clients that launch the server as a subprocess.

twenty-mcp --transport streamable-http --host 0.0.0.0 --port 8000

A network server with a /health endpoint and /mcp route.

twenty-mcp --transport sse --host 0.0.0.0 --port 8000

Health check (HTTP transports):

curl -s http://localhost:8000/health        # {"status":"OK"}

Configuration (environment)

twenty-mcp is configured entirely from the environment. The required set:

Var	Default	Meaning
TWENTY_URL	http://localhost:3000	Twenty CRM base server URL
TWENTY_TOKEN	twenty_developer_access_token	Developer access token (Bearer)
TWENTY_MCP_BASE_URL	http://localhost:3000/api	Base API URL to query
TWENTY_MCP_USERNAME	admin	Auth username (used if no token)
TWENTY_MCP_PASSWORD	secure_password	Auth password (used if no token)
TWENTY_TLS_PROFILE	system	Named outbound TLS policy from AgentConfig
TWENTY_API_PREFIX	/rest	API path prefix for record operations
CRMTOOL	True	Register the CRM tool
METADATATOOL	True	Register the metadata tool
OAUTHTOOL	True	Register the OAuth/webhooks tool

Plus HOST / PORT / TRANSPORT for HTTP transports. The full set is documented in .env.example. Copy it to .env and populate only what you use; an absent TWENTY_TOKEN leaves the client unauthenticated rather than raising.

Backing service

Twenty CRM is self-hostable — a Docker recipe to deploy the platform that TWENTY_URL points at is provided in Backing Platform. Twenty also offers a managed cloud instance; in that case only the connection configuration above is required.

Docker Compose

The repo ships docker/mcp.compose.yml. It reads a sibling .env and publishes the HTTP server on :8000:

services:
  twenty-mcp:
    image: example/twenty-mcp@sha256:<digest>
    container_name: twenty-mcp
    hostname: twenty-mcp
    restart: always
    env_file:
      - ../.env
    environment:
      - PYTHONUNBUFFERED=1
      - HOST=0.0.0.0
      - PORT=8000
      - TRANSPORT=streamable-http
      - TWENTY_URL
      - TWENTY_TOKEN
    ports:
      - "8000:8000"
    healthcheck:
      test: ["CMD", "python3", "-c", "import urllib.request; urllib.request.urlopen('http://localhost:8000/health')"]
      interval: 30s
      timeout: 10s
      retries: 3

cp .env.example .env          # then edit TWENTY_* values
docker compose -f docker/mcp.compose.yml up -d
docker compose -f docker/mcp.compose.yml logs -f

Agent server (A2A)

The A2A agent is published as the twenty-agent console script and docker/agent.compose.yml, which provisions the MCP server and the agent together. The agent reaches the MCP server over MCP_URL and publishes its own HTTP API (and optional web UI) on :9000:

services:
  twenty-mcp:
    image: example/twenty-mcp@sha256:<digest>
    hostname: twenty-mcp
    env_file: [../.env]
    environment:
      - TRANSPORT=streamable-http
      - HOST=0.0.0.0
      - PORT=8000
    ports: ["8000:8000"]

  twenty-agent:
    image: example/twenty-mcp@sha256:<digest>
    depends_on: [twenty-mcp]
    command: ["twenty-agent"]
    env_file: [../.env]
    environment:
      - HOST=0.0.0.0
      - PORT=9000
      - MCP_URL=http://twenty-mcp:8000/mcp
      - PROVIDER=${PROVIDER:-openai}
      - MODEL_ID=${MODEL_ID:-gpt-4o}
      - ENABLE_WEB_UI=True
    ports: ["9000:9000"]

docker compose -f docker/agent.compose.yml up -d
curl -s http://localhost:9000/health        # agent health

Configure the agent's model with PROVIDER, MODEL_ID, and the matching provider API key (for example LLM_API_KEY / OPENAI_API_KEY).

Behind a Caddy reverse proxy

Expose the HTTP server on a hostname with automatic TLS. Add to your Caddyfile:

# Internal (self-signed) — homelab .example.invalid zone
twenty-mcp.example.invalid {
    tls internal
    reverse_proxy twenty-mcp:8000
}

# Public — automatic Let's Encrypt
twenty-mcp.example.com {
    reverse_proxy twenty-mcp:8000
}

Reload Caddy:

docker compose -f services/caddy/compose.yml exec caddy caddy reload --config /etc/caddy/Caddyfile

DNS with Technitium

Point the hostname at the host running Caddy. Via the Technitium API:

curl -s "http://technitium.example.invalid:5380/api/zones/records/add" \
  --data-urlencode "token=$TECHNITIUM_DNS_TOKEN" \
  --data-urlencode "domain=twenty-mcp.example.invalid" \
  --data-urlencode "zone=arpa" \
  --data-urlencode "type=A" \
  --data-urlencode "ipAddress=192.0.2.10" \
  --data-urlencode "ttl=3600"

…or add an A record twenty-mcp.example.invalid → <caddy-host-ip> in the Technitium web console (http://technitium.example.invalid:5380). The ecosystem technitium-dns-mcp automates this as a tool.

Register with an MCP client

Add to your client's mcp_config.json:

{
  "mcpServers": {
    "twenty_mcp": {
      "command": "python",
      "args": ["-m", "twenty_mcp.mcp_server"],
      "env": {
        "TWENTY_URL": "http://your-twenty:3000",
        "TWENTY_TOKEN": "your_developer_access_token"
      }
    }
  }
}

For a remote HTTP server, point the client at http://twenty-mcp.example.invalid/mcp instead.