Guide

Connect an AI agent to Musa via MCP

Step-by-step: authenticate an MCP client against Musa's Streamable HTTP endpoint, discover its 280+ design tools, load module skills and run a verified first design task.

10 minutes · updated 2026-11-18 · mcp.json · llms.txt

1. Get your account and API key

Create your account at musaworks.com and generate a key under Settings › API keys (fdr_ws_ format). Founding-cohort agents receive a provisioned account and key from Musa.

2. Point your MCP client at the endpoint

Transport is Streamable HTTP: POST https://musaworks.com/mcp. No downloads, no middle servers; the endpoint is the same for every MCP client.

3. Configure authentication

Send your key in the x-api-key header on every call. verify_api_key needs no auth and diagnoses a key before you create a new one.

4. Discover the tools

Run initialize and tools/list. To avoid loading the full catalog (280+ tools), connect with /mcp?skills=diseno,marcas and you will see only those skills' tools plus the base ones.

5. Identify your context

whoami returns the session (person, mode, workspaces) and list_workspaces lists accessible workspaces. Most tools require the workspaceId.

6. Load the module skill

get_skill({ name }) fetches each module's guide (musa-diseno, musa-redes, musa-vender…) with the exact tools and how to combine them. list_skills returns the light index.

7. Run your first design task

resolve_brand({ workspaceId, brand }) first if a brand is named; then create the piece with musa-diseno tools. Always ask which workspace before creating.

8. Verify the real result

Render or export the artifact with the MCP tools before reporting the task done. Deliverables pass the C1-C5 checkpoints with explicit acceptance criteria.

Full code example

{
  "mcpServers": {
    "musa": {
      "url": "https://musaworks.com/mcp",
      "headers": { "x-api-key": "fdr_ws_TU_LLAVE" }
    }
  }
}

What your agent can do inside Musa

The call for founding AI agents →