search_apis

Instructions

Implementation Reference

• mcp/src/mcp/adapters/mcp.py:20-47 (handler)

  Core implementation of the search_apis tool handler in MCPAdapter. Constructs SearchRequest and calls Jentic SDK search method.

  async def search_api_capabilities(self, request: dict[str, Any]) -> dict[str, Any]:
      """MCP endpoint for searching API capabilities.
  
      Args:
          request: MCP tool request parameters.
  
      Returns:
          MCP tool response.
      """
      # Build SearchRequest using the new SDK
      search_request = SearchRequest(
          query=request.get("capability_description") or request.get("query", ""),
          keywords=request.get("keywords"),
          limit=request.get("max_results", 5),
          apis=request.get("api_names"),
      )
  
      search_response = await self.jentic.search(search_request)
  
      # We adopt the unified results list returned by SearchResponse
      response_data = search_response.model_dump(exclude_none=False)
      return {
          "result": {
              "matches": response_data,
              "query": search_request.query,
              "total_matches": search_response.total_count,
          }
      }

• mcp/src/mcp/tools.py:18-46 (schema)

  JSON schema definition for the search_apis tool, defining input parameters and structure.

  SEARCH_API_CAPABILITIES_TOOL = {
      "name": "search_apis",
      "description": "Search for available actions or information based on what the user wants to do (e.g., 'find Discord servers', 'send a message'). Use this first to understand what's possible.",
      "parameters": {
          "type": "object",
          "properties": {
              "capability_description": {
                  "type": "string",
                  "description": "Natural language description of the action needed (e.g., 'send emails', 'weather forecasting', 'natural language processing')",
              },
              "keywords": {
                  "type": "array",
                  "description": "Optional list of specific keywords to help narrow down the search",
                  "items": {"type": "string"},
              },
              "max_results": {
                  "type": "integer",
                  "description": "Maximum number of actions to return",
                  "default": 5,
              },
              "api_names": {
                  "type": "array",
                  "description": "Optional list of API names to restrict results to specific APIs. Use vendor format (e.g., 'google.com') or sub-API format (e.g., 'atlassian.com/jira'). Only results from these APIs will be shown, regardless of search query.",
                  "items": {"type": "string"},
              },
          },
          "required": ["capability_description"],
      },
  }

• mcp/src/mcp/handlers.py:45-50 (registration)

  Registration of search_apis handler in the general request handler mapping.

  tool_handlers = {
      "search_apis": mcp_adapter.search_api_capabilities,
      "load_execution_info": mcp_adapter.generate_runtime_config,
      "execute": mcp_adapter.execute,  # Add the execute tool handler
      "submit_feedback": mcp_adapter.submit_feedback
  }

• mcp/src/mcp/transport/stdio.py:36-42 (registration)

  Registration of search_apis in stdio transport handlers dictionary.

  self._handlers = {
      "search_apis": self._handle_search_api_capabilities,
      "load_execution_info": self._handle_generate_runtime_from_selection_set,
      "generate_code_sample": self._handle_generate_code_sample,
      "execute": self._handle_execute,  # Add execute handler
      "submit_feedback": self._handle_submit_feedback, # Add submit_feedback handler
  }

• mcp/src/mcp/transport/http.py:76-81 (registration)

  HTTP endpoint registration for search_apis in HTTP transport, delegating to adapter.

  @self._app.post("/api/search_apis")
  async def search_api_capabilities(request: Request):
      """MCP endpoint for searching API capabilities."""
      data = await request.json()
      result = await self.adapter.search_api_capabilities(data)
      return JSONResponse(result)