CreateWatch
Sign inStart creating

Agent kit — portable bundles for Claude Code, Cursor, OpenAI Agents, LangChain, n8n

Download a zip for your framework: MCP config, the tool catalogue as JSON, one SKILL.md per production skill, .env example and a five-minute README. No key inside; paste your own.

Open it →

Agent kit (top nav) packages everything an agent framework needs to drive this studio, in that framework's layout:

  • Claude Code / Claude Desktop — .mcp.json, claude_desktop_config.json, a CLAUDE.md, and skills/<name>/SKILL.md for each production skill.
  • Cursor / Windsurf — .cursor/mcp.json and a rules file.
  • OpenAI Agents SDK, LangChain / LangGraph — a Python starter that registers the MCP server.
  • n8n — an importable workflow with an MCP Client node.
  • Any HTTP client — the JSON-RPC calls spelled out in a .http file.

Every kit also has tools.json (generated from the live server — if it ever disagrees with the server, the server is right), .env.example, and a README that says what to do in the first five minutes.

Nothing in the bundle is a secret. Create an API key under Settings → MCP, scope it to a project if the agent only needs one, and paste it in. The key's scope and spend cap are enforced by the server, not by the agent's good behaviour.

Did this answer your question?

Related

Flows — save a chain of tools and re-run it

Build a recipe once — reference pack, video, social cuts, captions — then run it on any new input.

All help articlesAsk the community