ContextEngine MCP Server

Instructions

Implementation Reference

• src/lib/api.ts:91-155 (handler)

  Core implementation of the startContextEngine function, which handles local documentation setup and remote API call to start the context engine.

  export async function startContextEngine(
    projectRoot: string,
    clientIp?: string,
    apiKey?: string,
    serverUrl?: string
  ): Promise<string> {
    if (!projectRoot) {
      throw new Error("Project root directory is required for context engine initialization");
    }
  
    let apiResponse = "";
    let documentationStatus = "";
  
    try {
      // Step 1: Setup local documentation structure first
      logger.info("Setting up local documentation structure");
      const setupResult = await setupDocumentationStructure(projectRoot);
  
      if (setupResult.success) {
        documentationStatus = `\n\n📁 Local Documentation Structure: ${setupResult.message}`;
        logger.info("Local documentation structure setup completed", {
          status: setupResult.status as unknown as Record<string, unknown>,
        });
      } else {
        logger.warn("Local documentation structure setup failed", {
          status: setupResult.status as unknown as Record<string, unknown>,
        });
        // Don't proceed with API call if local setup fails
        return `❌ Failed to setup local documentation structure: ${setupResult.message}`;
      }
    } catch (error) {
      logger.error("Local documentation structure setup failed", {
        error: error instanceof Error ? error.message : String(error),
      });
      // Don't proceed with API call if local setup fails
      return `❌ Failed to setup local documentation structure: ${error instanceof Error ? error.message : String(error)}`;
    }
  
    try {
      // Step 2: Start the context engine via API (only if local setup succeeded)
      logger.info("Starting ContextEngine via API");
      const url = buildApiUrl("start-context-engine", {}, serverUrl);
      const headers = generateHeaders(clientIp, apiKey, { "X-ContextEngine-Source": "mcp-server" });
  
      const response = await makeApiRequest(url, headers, "start context engine");
      const text = await response.text();
  
      if (!text || text === "No content available") {
        apiResponse = "Context engine start request sent but no confirmation available.";
      } else {
        apiResponse = text;
      }
  
      logger.info("ContextEngine API call successful");
    } catch (error) {
      logger.error("ContextEngine API call failed", {
        error: error instanceof Error ? error.message : String(error),
      });
  
      apiResponse = `⚠️  ContextEngine API call failed: ${error instanceof Error ? error.message : String(error)}`;
    }
    
    // Return combined response
    return `${apiResponse}${documentationStatus}`;
  }

• src/index.ts:77-107 (registration)

  MCP tool registration for 'start_context_engine', including input schema, description, and handler that delegates to the core startContextEngine function.

  server.registerTool(
    "start_context_engine",
    {
      title: "Start Context Engine",
      description: "Starts the context engine and returns a confirmation message.",
      inputSchema: {
        projectRoot: z.string().describe("The project root directory"),
      },
    },
    async (args) => {
      // Validate API key based on environment
      validateApiKey(apiKey, finalServerUrl);
  
      const startContextEngineResponse = await startContextEngine(
        args.projectRoot,
        clientIp,
        apiKey,
        finalServerUrl
      );
  
      return {
        content: [
          {
            type: "text",
            text: startContextEngineResponse,
          },
        ],
      };
    }
  );

• src/index.ts:80-86 (schema)

  Input schema definition for the tool using Zod, requiring projectRoot string.

  {
    title: "Start Context Engine",
    description: "Starts the context engine and returns a confirmation message.",
    inputSchema: {
      projectRoot: z.string().describe("The project root directory"),
    },
  },