git clone <repo-url>
cd mcpdocs
npm install

Create a .env file in the root with the following variables:

LAMATIC_ENDPOINT=https://your-project.lamatic.dev/graphql
LAMATIC_API_KEY=your_api_key
LAMATIC_PROJECT_ID=your_project_id
LAMATIC_WORKFLOW_ID=your_workflow_id

# Optional — raises the GitHub API rate limit for kit_check_pr_status, and is
# required for kit_revalidate_pr (stdio only) to post comments. On the hosted
# deployment (docmcp.lamatic.ai) this must be set as a Vercel project env var.
GITHUB_TOKEN=your_github_personal_access_token

Tool definitions live in tools/registry.js, shared by both entry points below.

• query_docs — RAG search over Lamatic.ai docs.
• kit_list, kit_get, kit_search, kit_get_flow — browse the AgentKit repo, no auth required.
• kit_check_pr_status — read-only AgentKit PR validation status (uses GITHUB_TOKEN if set, for a higher rate limit).
• kit_validate_structure, kit_auth_login, kit_revalidate_pr — stdio only. kit_validate_structure needs the local filesystem, kit_auth_login persists a token to local disk (~/.lamatic/config.json), and kit_revalidate_pr is a write action — the hosted endpoint has no caller authentication, so exposing a write tool there would let any anonymous caller post GitHub comments through the server's shared token.

To test the local stdio MCP, use the command below:

git clone https://github.com/Lamatic/Lamatic-MCP-Docs
cd Lamatic-MCP-Docs
npm install
cp .env.example .env  # fill in your Lamatic credentials
npm run dev

• Lamatic.ai
• Lamatic Docs
• MCP Endpoint
• Live Homepage