ChatGPT MCP Server

Instructions

Implementation Reference

• src/servers/mcp.server.ts:145-158 (registration)

  Registration of the 'container_start' tool including its name, description, and input schema definition.

  {
    name: 'container_start',
    description: 'Start a stopped container',
    inputSchema: {
      type: 'object',
      properties: {
        container: {
          type: 'string',
          description: 'Container ID or name',
        },
      },
      required: ['container'],
    },
  },

• src/servers/mcp.server.ts:257-263 (handler)

  Handler logic for the 'container_start' tool: parses arguments, calls DockerService.startContainer, and formats the response.

  case 'container_start': {
    const { container } = request.params.arguments as { container: string };
    const output = await this.dockerService.startContainer(container);
    return {
      content: [{ type: 'text', text: `Container started: ${output}` }],
    };
  }

• src/services/docker.service.ts:49-51 (helper)

  Core implementation of container start: executes the Docker CLI command 'docker start <container-id>'.

  async startContainer(id: string): Promise<string> {
    return this.executeCommand(`start ${id}`);
  }

• src/services/docker.service.ts:8-18 (helper)

  Utility method to execute Docker CLI commands asynchronously, used by all Docker operations including startContainer.

  async executeCommand(command: string): Promise<string> {
    try {
      const { stdout } = await execAsync(`docker ${command}`);
      return stdout;
    } catch (error: any) {
      throw new McpError(
        ErrorCode.InternalError,
        `Docker command failed: ${error.message}`
      );
    }
  }