Skip to main content
These MCP tools are the registry query API for agents. Use them when your agent needs grounded tool discovery at runtime instead of recalled package names or guessed integrations. Prefer the REST API if you are integrating outside an MCP client.

​
Server URL

All MCP tools are available at:
Querying UseAgents via MCP is currently free during our public beta. No API key is required. See Authentication for details.

​
Response format

MCP tools default to Toon (Token-Oriented Object Notation), a compact, token-efficient format designed for LLM consumption. Pass format: "json" on any tool call to get JSON text instead. Learn more at toonformat.dev. See Response formats for API and CLI defaults.

​
What the tools do

Agents use these tools to:
  • Search the registry by natural-language problem statement and constraints
  • Shortlist candidate tools with compact registry metadata
  • Retrieve authoritative bootstrap context (install, examples, I/O) for a selected tool
  • Search that tool’s official documentation for deeper how-to answers
  • Optionally run a small program in a UseAgents sandbox with test_tool
  • Hand off to official docs for full API depth instead of guessing from training data

​
Current reference pages

Query the registry in three steps:
  1. Call search_tools to find candidate tools by task, framework, or capability.
  2. Call get_tool_context with the selected slug before recommending installation steps, code, or usage patterns.
  3. Call search_docs with the same slug and a specific question when you need deeper answers from that tool’s official docs.
  4. Optionally call test_tool with the files you wrote. Pass env when the snippet needs config or secrets.
If get_tool_context has no result for a slug, treat the tool context as unavailable instead of guessing from training data.

​
Next steps