Enterprise_DevOps_MCP

Bundles GitHub, CircleCI, Grafana, Slack, Kubernetes, and 7 custom tools into a single E2B sandbox with a sub-2-second cold start. ## Prerequisites - Node.js 18+ - E2B account and `E2B_API_KEY` - API tokens for connected services: `GITHUB_TOKEN`, `CIRCLECI_TOKEN`, `GRAFANA_API_KEY`, `GRAFANA_URL`, `SLACK_BOT_TOKEN` ## Setup **Step 1: Clone and install dependencies** ```bash git clone https://github.com/CoderRahul01/Enterprise_DevOps_MCP.git cd Enterprise_DevOps_MCP npm install ``` *Expected Output:* ``` added 42 packages, and audited 43 packages in 3s found 0 vulnerabilities ``` **Step 2: Configure environment variables** ```bash cp .env.example .env ``` *Expected Output:* ``` (No output) ``` **Step 3: Build the custom MCP server** ```bash npm run mcp-server:build ``` *Expected Output:* ``` > devops-agent-mcp@1.0.0 mcp-server:build > cd custom-mcp-server && npm run build > internal-devops-mcp@1.0.0 build > tsc ``` **Step 4: Build the production E2B template** ```bash npm run build:prod ``` *Expected Output:* ``` > devops-agent-mcp@1.0.0 build:prod > tsx scripts/build.prod.ts Building E2B template... ✅ Snapshot built successfully Template ID: devops-mcp-prod-xyz123 ``` ## Run the Demo To test the unified endpoint locally, run the development script. ```bash npm run build:dev ``` *Expected Output:* ``` Starting E2B sandbox... ✅ Sandbox ready Connecting to MCP servers... MCP connection successfully established. ``` ## Run the AI Agent Initialize your agent connecting to the unified endpoint. Use these example prompts to verify all toolsets are accessible: - "Check the deployment history for the frontend service." - "Get the service health for the payment gateway. If errors are above 5%, notify the on-call engineer in Slack." - "View the logs for the `auth-pod` in Kubernetes and summarize any errors." ## Add Internal APIs Open `custom-mcp-server/src/index.ts`. First, define your new tool schema inside the `ListToolsRequestSchema` handler. Second, define the execution logic inside the `CallToolRequestSchema` switch statement using standard TypeScript. Finally, run `npm run mcp-server:build` and `npm run build:prod` to bake the new tool into your E2B snapshot. ## Output Usage When the sandbox boots in your application code, it returns the connection credentials. *Expected Output:* ``` mcpUrl: https://mcps.e2b.dev/sandbox/devops-mcp-prod-xyz123 mcpToken: e2b_sk_example123456789 ``` Pass the `mcpUrl` and `mcpToken` directly to the MCP client configuration in frameworks like the Anthropic SDK or LangGraph to bind the tools to an LLM.