Mesh Agent MCP Server

Instructions

Implementation Reference

• mesh_mcp_server/tool_server.py:392-409 (handler)

  Handler function that executes the tool by looking up agent and tool name from registry and calling the Mesh API execute_tool method. This handles 'coingeckotokeninfoagent_get_trending_coins'.

  async def call_tool(name: str, arguments: dict) -> List[types.TextContent]:
      """Call the specified tool with the given arguments."""
      try:
          if name not in self.tool_registry:
              raise ValueError(f"Unknown tool: {name}")
  
          tool_info = self.tool_registry[name]
          result = await self.execute_tool(
              agent_id=tool_info["agent_id"],
              tool_name=tool_info["tool_name"],
              tool_arguments=arguments,
          )
  
          # Convert result to TextContent
          return [types.TextContent(type="text", text=str(result))]
      except Exception as e:
          logger.error(f"Error calling tool {name}: {e}")
          raise ValueError(f"Failed to call tool {name}: {str(e)}") from e

• mesh_mcp_server/tool_server.py:264-301 (registration)

  Code that dynamically registers tools by constructing tool_id as '{agent_id.lower()}_{tool_name}', loading schema from metadata, and storing in registry. This creates the 'coingeckotokeninfoagent_get_trending_coins' tool if configured.

  for tool in agent_data.get("tools", []):
      if tool.get("type") == "function":
          function_data = tool.get("function", {})
          tool_name = function_data.get("name")
  
          if not tool_name:
              continue
  
          # Check if this tool is enabled based on configuration
          if not self.is_tool_enabled(agent_id, tool_name):
              logger.debug(
                  f"Skipping tool {tool_name} for agent {agent_id} (not in config)"  # noqa: E501
              )
              tools_skipped += 1
              continue
  
          # Create a unique tool ID
          tool_id = f"{agent_id.lower()}_{tool_name}"
  
          # Get parameters or create default schema
          parameters = function_data.get("parameters", {})
          if not parameters:
              parameters = {
                  "type": "object",
                  "properties": {},
                  "required": [],
              }
  
          # Store tool info
          tool_registry[tool_id] = {
              "agent_id": agent_id,
              "tool_name": tool_name,
              "description": function_data.get("description", ""),
              "parameters": parameters,
          }
          tools_enabled += 1
          logger.debug(f"Enabled tool: {tool_id}")

• mesh_mcp_server/tool_server.py:316-353 (helper)

  Helper function that performs the actual API call to the Mesh sequencer to execute the agent's tool.

      self, agent_id: str, tool_name: str, tool_arguments: Dict[str, Any]
  ) -> Dict[str, Any]:
      """Execute a tool on a mesh agent.
  
      Args:
          agent_id: ID of the agent to execute the tool on
          tool_name: Name of the tool to execute
          tool_arguments: Arguments to pass to the tool
  
      Returns:
          Tool execution result
  
      Raises:
          ToolExecutionError: If there's an error executing the tool
      """
      request_data = {
          "agent_id": agent_id,
          "input": {"tool": tool_name, "tool_arguments": tool_arguments},
      }
  
      # Add API key if available
      if Config.HEURIST_API_KEY:
          request_data["api_key"] = Config.HEURIST_API_KEY
  
      try:
          result = await call_mesh_api(
              "mesh_request", method="POST", json=request_data
          )
          return result.get("data", result)  # Prefer the 'data' field if it exists
      except MeshApiError as e:
          # Re-raise API errors with clearer context
          raise ToolExecutionError(str(e)) from e
      except Exception as e:
          logger.error(f"Error calling {agent_id} tool {tool_name}: {e}")
          raise ToolExecutionError(
              f"Failed to call {agent_id} tool {tool_name}: {str(e)}"
          ) from e

• mesh_mcp_server/server.py:40-52 (registration)

  In server.py variant, the CoinGeckoTokenInfoAgent is explicitly listed as a supported agent, enabling its tools like get_trending_coins to be dynamically registered as coingeckotokeninfoagent_get_trending_coins.

  DEFAULT_AGENTS = [
      "CoinGeckoTokenInfoAgent",
      "DexScreenerTokenInfoAgent",
      "ElfaTwitterIntelligenceAgent",
      "ExaSearchAgent",
      "TwitterInfoAgent",
      "AIXBTProjectInfoAgent",
      "EtherscanAgent",
      "EvmTokenInfoAgent",
      "FundingRateAgent",
      "UnifaiTokenAnalysisAgent",
      "YahooFinanceAgent",
      "ZerionWalletAnalysisAgent"

• mesh_mcp_server/tool_server.py:379-389 (registration)

  MCP list_tools handler that exposes the dynamically registered tools, including their schemas.

  @app.list_tools()
  async def list_tools() -> List[types.Tool]:
      """List all available tools."""
      return [
          types.Tool(
              name=tool_id,
              description=tool_info["description"],
              inputSchema=tool_info["parameters"],
          )
          for tool_id, tool_info in self.tool_registry.items()
      ]