@lamatic/graph-mcp

Execute your deployed Lamatic flows via AI agents using natural language.

What is it?

Lamatic GraphMCP is an MCP (Model Context Protocol) server that turns your deployed Lamatic flows into AI-callable tools. Connect Claude, GitHub Copilot, or Cursor — your flows become tools your AI can call directly using natural language.

Installation

# Run directly via npx (recommended)
npx @lamatic/graph-mcp

# Or install globally
npm install -g @lamatic/graph-mcp

# Or install as a dependency
npm install @lamatic/graph-mcp

Setup

Claude Code

claude mcp add lamatic-graph -- npx -y @lamatic/graph-mcp

VS Code

Add to .vscode/mcp.json:

{
  "servers": {
    "lamatic-graph": {
      "command": "npx",
      "args": ["-y", "@lamatic/graph-mcp"],
      "type": "stdio"
    }
  }
}

Claude Desktop / Cursor

Add to your MCP config file:

{
  "mcpServers": {
    "lamatic-graph": {
      "command": "npx",
      "args": ["-y", "@lamatic/graph-mcp"]
    }
  }
}

Authentication

GraphMCP requires two API keys — an org-level key and a project-level key.

Run the graph_auth_login tool:

authenticate with orgApiKey lt-org-xxx, projectApiKey lt-mmup-xxx and orgId your-org-id

Your credentials are stored locally at ~/.lamatic/config.json.

Tools

Tool	Description
graph_auth_login	Authenticate with Lamatic
graph_load_project_flows	Load all active flows with input schemas
graph_execute_flow	Execute a deployed flow with a payload
graph_refresh_flows	Refresh the list of active flows
graph_generate_flow_skills	Generate one Claude Skill per flow (what it does / when / how to call)

Skill (recommended)

GraphMCP ships a Claude Agent Skill at skills/lamatic-flows/ that teaches Claude the full "do a task for me" workflow: discover flows, pick the right one, build a schema-matched payload, execute, and report the result.

The MCP gives Claude the tools; the skill gives it the procedure — so behavior stays reliable instead of Claude re-figuring-out the flow each time.

To use it, copy the skill into your Claude Code skills directory:

cp -r node_modules/@lamatic/graph-mcp/skills/lamatic-flows ~/.claude/skills/
# or, for a single project: .claude/skills/

Then just ask in natural language ("run the Changelog flow with …") and Claude will follow the skill.

Per-flow skills (generated)

Beyond the generic skill, you can generate one skill per flow so each flow becomes its own discoverable tool. Run the graph_generate_flow_skills tool (full mode / org access required):

generate skills for my flows in project fcd1aa1d-...

For every active flow it writes a SKILL.md under ./.claude/skills/lamatic-<flow-name>/ describing:

• What the tool can do — the flow's purpose
• When to call it — routes the user's intent to this flow (via the skill description)
• How to call it — the exact input fields + an example graph_execute_flow call

Reload skills in your client afterwards to pick them up.

Supported Flow Triggers

GraphMCP can execute flows with the following trigger types:

Trigger	Supported
API Request	✅
Chat Widget	✅
Search Widget	✅
Webhook	✅
Cron	⚠️ Requires trigger data
Google Sheets	⚠️ Requires trigger data

How It Works

1. Connect your MCP client with org API key + project API key + org ID
2. Run graph_load_project_flows to see all active flows and their input fields
3. Run graph_execute_flow with the flow ID and matching payload
4. The AI maps your natural language to the right flow and payload automatically

Example Usage

show me all flows for project fcd1aa1d-...
run the Changelog flow with sampleInput "hello world"
execute the RAG flow and ask it "what is Lamatic?"
refresh my flows

Links

• Lamatic
• DevMCP — Manage Lamatic projects via AI
• npm