last9

Last9 MCP Server exposes a local MCP transport that lets agents like Claude, Cursor, Windsurf, and VSCode integrate real-time production context into your environment. By running the local STDIO server (via npx), you can connect your MCP-enabled agent to the Last9 API and query observability data, logs, traces, and change events through the server’s provided tools. The server supports multiple MCP tools, including Observability & APM, Prometheus/PromQL, Logs Management, Traces Management, Change Events, and Alerts management, enabling you to fetch service summaries, metrics, logs, traces, and alert configurations in a unified interface. Use the standard MCP client configuration to point your agent to the local server and, if desired, fall back to the hosted HTTP transport for a managed experience.

To start using it, install/run the MCP server via npm/npx as shown in installation instructions, then configure your MCP client with the server name (e.g., last9) and the appropriate transport. If you prefer not to run a local binary, you can use the hosted HTTP transport by configuring the client with the Last9 HTTP endpoint and API token as documented in the README.

Prerequisites:\n- Node.js and npm/yarn installed on your machine.\n- Access to the internet to install the MCP server package via npm/yarn.\n\nInstall (recommended: managed HTTP transport)\n- If you want to use the hosted HTTP transport, skip local installation and configure your MCP client with the Last9 HTTP URL and token as shown in the README.\n\nLocal STDIO fallback (if your client requires a local binary):\n1) Install the MCP server globally via npm:\nbash\nnpm install -g @last9/mcp-server@latest\n\n2) Run the server directly (STDIO):\nbash\nlast9-mcp-server --version\n# Or rely on npx as shown in the configuration example:\nnpx -y @last9/mcp-server@latest\n\n3) Point your MCP client to the local STDIO server (the specific path/descriptor is defined by your client’s MCP configuration).\n\nAlternative installation (GitHub releases)\n- Download the appropriate binary for your platform from GitHub Releases and run it locally, then configure your MCP client to use the local binary path.\n\nPrerequisites recap:\n- Node.js (recommended) or a compatible runtime for the MCP server package.\n- Optional: access to Last9 credentials if using the HTTP transport.\n

Tips and caveats:\n- If you choose the managed HTTP transport, you’ll need a Client Token with MCP permissions and your organization slug. This setup routes requests through Last9’s hosted API rather than a local STDIO process.\n- The MCP server loads and caches log/trace attribute names at startup and refreshes periodically; this helps AI agents construct up-to-date queries.\n- For absolute time queries, prefer RFC3339/ISO8601 timestamps (start_time_iso, end_time_iso) and use lookback_minutes for relative windows if supported by the tool.\n- The server exposes a wide range of tools (observability, Prometheus queries, logs, traces, change events, and alerts). Review the Tools Documentation section of the README to understand parameters and response formats for each tool.\n- When using the local binary, ensure your environment has network access to Last9 services if you’re combining local and hosted transports.\n- If you encounter time window or parameter validation errors, verify your input times and consider using explicit start_time_iso/end_time_iso values.\n