MCP server

Point an MCP client at Mellow AI. Agents can read who we are, and get a model-specific rewrite of a prompt that is not landing.

Connect

Streamable HTTP. Cursor, Claude Code, or any MCP client.

mcp.json

{
  "mcpServers": {
    "mellow-ai": {
      "url": "https://staymellow.ai/api/mcp"
    }
  }
}

Claude Code: claude mcp add --transport http mellow-ai https://staymellow.ai/api/mcp

Two tools, six resources.

get_mellow_info
Studio facts — who we are, the work, how we engage, contact, newsletter.
optimize_prompt
Rewrite a prompt for Claude, GPT, Gemini, Grok, Llama, or another named model.
mellow://about
About
mellow://work
Work
mellow://process
Process
mellow://contact
Contact
mellow://newsletter
Signal
mellow://prompting/{family}
Model guides

Prompt optimization uses the target model's family guide, then rewrites the draft. Pass the model id you will actually run, not a generic "LLM".

Plain text versions: llms.txt and llms-full.txt