MCP Tools Reference
ToolPilot exposes 5 MCP tools that AI agents use to discover, evaluate, and report on developer tools. Every tool follows the Model Context Protocol specification and can be called by any MCP-compatible client.
All tools
search_toolsDiscoveryPrimary discovery tool. Accepts a natural-language query and returns tool recommendations through a multi-stage pipeline.
search_tools_respondDiscoveryAnswers clarification questions returned by search_tools to refine and complete a search.
get_stackStackBuilds a compatible tool stack for a given use case using graph relationship data.
check_issueDiagnosticsDiagnoses known issues with a specific tool and returns matched solutions with confidence scores.
report_outcomeFeedbackReports the outcome of using a recommended tool so the graph can improve future recommendations.
Typical flow
A standard agent interaction follows a three-step pattern:
- 1
search_toolsThe agent sends a natural-language query. ToolPilot either returns results immediately or asks clarification questions.
- 2
search_tools_respond(optional)If clarification was needed, the agent answers the questions. ToolPilot completes the pipeline and returns results.
- 3
report_outcomeAfter the agent uses a recommended tool, it reports success, failure, or partial success. This feedback improves future recommendations.
MCP Protocol Compatibility
Supplementary tools
get_stack and check_issue are standalone tools that can be called independently of the search flow. Use get_stack when building a new project and check_issue when troubleshooting an existing tool.