Sushi Tools

SushiMCP comes loaded with a set of tools designed to feed context to your AI IDE's. Not just any context, but up to date, accurate, context from the source documentation.

Available Tools

• list_llms_txt_sources

  Lists the names of all configured documentation sources.

  Input:

  None.

  Output:

  A JSON array containing the names of the configured sources (strings).

  ["drizzle", "hono", "mylocal"]
• fetch_llms_txt

  Fetches the content of a specific documentation source by its configured name.

  Input:

  A JSON object specifying a single source by its configured name:{"url": "drizzle"}

  Or, an array of configured source names:["drizzle", "hono"]

  Output:

  The text content of the fetched source(s).

  Security:

  The server will enforce the --allow-domain rules for remote URLs.

Roadmap

Best Practices

Best practices when requesting your AI agent to call SushiMCP tools:

• Carefully consider how you make requests for documentation. Base models sometimes fall into a loop when missing documentation is requested in a way that leads the model to believe the documentation should be available.

Ask SushiMCP for XYZ docs

--vs--

Ask SushiMCP for any available documentation for Hono and Drizzle.

• Remember that some projects make llms-full.txt and llms-mini.txt available alongside llms.txt
• Some llms.txt are just a collection of links - request an llms-full.txt or llms-mini.txt if available
• Request the appropriate level of documentation based on your needs
• Take into account the context window made available by your chosen model
• The llms.txt standard is relatively new so expect that many projects will not adhere

---

Tool Request

Have a tool idea you want added? Submit a request