mcp-server-logs-sieve is a Model Context Protocol (MCP) server that lets AI assistants query logs directly from observability backends. Its goal is to ask debugging questions in natural language, with Logs Sieve pulling the exact logs and context.
Purpose During incident handling, a lot of time is spent on repetitive steps: finding the right log source, narrowing the time window, discovering recurring failure patterns, and tracing a request across services. Logs Sieve packages these workflows into tools for AI assistants to call.
Main capabilities The server provides four tools: query_logs, summarize_logs, trace_request, and list_log_sources.
Supported providers
- Google Cloud Logging (gcp)
- AWS CloudWatch Logs (aws)
- Azure Monitor Logs (azure)
- Grafana Loki (loki)
- Elasticsearch (elasticsearch)
Usage Running requires Node.js 20 or later. The package uses ES modules and dependencies that require Node.js 20+; if you see an ERR_MODULE_NOT_FOUND error, check the Node version with node -v and upgrade as needed. This requirement applies whether running via npx, a global install, or from source.
Quick MCP configuration An example .mcp.json is available. If running directly from this repository's source, you can still use node ./bin/mcp-server-logs-sieve.js --provider gcp.
First-time setup: on first use, npx needs to download and install the package (about 40 seconds). Claude Code may time out while waiting. It is recommended to run it once in the terminal to warm the cache before adding it to the MCP configuration; afterward, Claude Code will start the server immediately from the npx cache. After updating the configuration, restart the MCP client and confirm with /mcp.
Provider authentication and environment variables The material mentions setup and authentication documentation for each provider, as well as Elasticsearch API compatibility headers. The package also includes a CLI for direct terminal use.
Getting started The material includes entries such as MCP client setup (Claude Code, Claude Desktop, Cursor, Windsurf), troubleshooting, and a compatibility matrix.
Contributing Contributions are welcome. The starting point is CONTRIBUTING.md, and the issue reporting address is https://github.com/Oluwatunmise-olat/mcp-server-logs-sieve/issues，讨论地址为 https://github.com/Oluwatunmise-olat/mcp-server-logs-sieve/discussions。
Limitations The material does not provide the specific parameters of the four tools, detailed authentication steps for each provider, the specific CLI commands, the specific contents of the compatibility matrix, or any fee or pricing information; these aspects are not currently provided.
