| name | lamatic |
| description | Operating guide for AI agents working with Lamatic.ai — a managed platform for building, deploying, and monitoring AI applications and agents. |
| version | 1.0.0 |
Lamatic Agent Skill
This document tells AI agents (Claude, Cursor, custom assistants, etc.) how to work effectively with Lamatic.ai. It is not general user documentation. For that, see llms.txt or the full docs.
What Lamatic is
Lamatic.ai is a managed platform for building AI-powered applications. The core unit of work is a Flow, a graph of connected Nodes that runs in order. A Flow is built in the Studio editor, deployed to an endpoint, and then called via API, SDK, CLI, or an MCP server.
Key concepts:
- Flow — a runnable graph of nodes. Triggered via API, schedule, webhook, widget, etc.
- Node — a single step in a flow (AI generation, data lookup, branching logic, integration call).
- Agent — a specialised AI node (Supervisor, Text, JSON, etc.) with reasoning capabilities.
- Context — vector and memory stores attached to a project for RAG and stateful interactions.
- Project — the deployment boundary. Each project has its own credentials, secrets, and deployment.
- Organization — owns one or more projects. Org-level operations require an Enterprise API key.
How to interact with Lamatic from outside Studio
Pick the right tool for the job:
| Goal | Use |
|---|
| Answer a question about Lamatic itself | Docs MCP — hosted, no auth needed |
| Run a deployed flow from natural language | Graph MCP |
| Manage projects / flows / credentials programmatically | CLI or Enterprise API |
| Call a flow from application code | SDK (JS/TS, React, Next.js, Go) |
| Make raw GraphQL requests | API Integration |
| Connect Lamatic flows to external MCP servers | MCP Node |
Authentication
Two key types:
- Project API key — scoped to one project. Used to trigger flows from an app. Header:
Authorization: Bearer <key> plus x-project-id: <project-id>.
- Enterprise API key — scoped to the whole organization (prefix
lt-org-...). Used for managing org-level resources. Required by the CLI, Graph MCP, and Dev MCP.
Never log, print, or commit either key. If a user pastes one in chat, treat it as sensitive.
Endpoints worth knowing
https://lamatic.ai/llms.txt — curated documentation index, markdown links.
https://lamatic.ai/llms-full.txt — full concatenated docs in one file.
https://lamatic.ai/<doc-url>.md — markdown version of any doc page (append .md).
https://docmcp.lamatic.ai/api/mcp — Docs MCP server (free, no auth).
https://<project>.lamatic.dev/graphql — per-project GraphQL endpoint for running flows.
Recommended behaviours
- Prefer the Docs MCP for any "how do I…" or "what does X mean" question. It answers from the live indexed docs.
- Prefer
.md URLs when fetching docs pages programmatically; you get clean markdown instead of rendered HTML.
- Use the CLI for repeatable org actions (deploy, create flow, update credentials). Avoid scripting raw GraphQL for management tasks unless the CLI doesn't cover it.
- Confirm destructive operations (deleting a project, deleting a flow, deploying to production) with the user before running them.
- Stay scoped to a project unless the user explicitly asks for org-level work. Most flow operations require only a Project API key.
- Quote field names verbatim from the docs when constructing payloads. Node parameter names use
snake_case in the low-code config (e.g. recipient_email, is_html, add_label_ids).
What to avoid
- Inventing endpoint paths, header names, or field names. If unsure, query the Docs MCP or read the relevant
.md page.
- Running
lamatic project delete or dev_delete_project without explicit user confirmation.
- Calling org-level endpoints with a Project API key (they will fail with an auth error).
- Treating output from the Docs MCP as authoritative for your own private flows — it indexes public docs only.
Source of truth
If something on this page contradicts the latest docs, the docs win.