railway-mcp

Instructions

Implementation Reference

• src/tools/deployment.tool.ts:94-120 (registration)

  Tool registration for 'deployment_status', including description, input schema, and handler function that delegates to deploymentService.healthCheckDeployment.

  createTool(
    "deployment_status",
    formatToolDescription({
      type: 'API',
      description: "Check the current status of a deployment",
      bestFor: [
        "Monitoring deployment progress",
        "Verifying successful deployments",
        "Checking for deployment failures"
      ],
      notFor: [
        "Service runtime logs",
        "Database logs"
      ],
      relations: {
        prerequisites: ["deployment_list", "deployment_trigger"],
        nextSteps: ["deployment_logs"],
        related: ["service_info", "service_restart", "deployment_wait"]
      }
    }),
    {
      deploymentId: z.string().describe("ID of the deployment to check status for")
    },
    async ({ deploymentId }) => {
      return deploymentService.healthCheckDeployment(deploymentId);
    }
  )

• src/tools/deployment.tool.ts:117-119 (handler)

  The handler function that executes the tool logic by calling the deployment service's healthCheckDeployment method.

  async ({ deploymentId }) => {
    return deploymentService.healthCheckDeployment(deploymentId);
  }

• src/tools/deployment.tool.ts:114-116 (schema)

  Zod input schema for the deployment_status tool, requiring a deploymentId string.

  {
    deploymentId: z.string().describe("ID of the deployment to check status for")
  },

• src/tools/index.ts:32-36 (registration)

  Global registration of all tools, including deployment_status, to the MCP server.

  allTools.forEach((tool) => {
    server.tool(
      ...tool
    );
  });

• src/services/deployment.service.ts:103-119 (helper)

  Supporting service method implementing the core deployment status check logic, with delay, API call, and formatted response.

  async healthCheckDeployment(deploymentId: string) {
    try {
      // Wait for 5 seconds before checking status
      // Seems like the LLMs like to call this function multiple times in combination
      // with the health check function, so we need to wait a bit
      await new Promise(resolve => setTimeout(resolve, 5000));
      const status = await this.client.deployments.healthCheckDeployment(deploymentId);
      const emoji = getStatusEmoji(status);
      
      return createSuccessResponse({
        text: `Deployment Status: ${emoji} ${status}`,
        data: { status }
      });
    } catch (error) {
      return createErrorResponse(`Error checking deployment health: ${formatError(error)}`);
    }
  }