Mcp Observability
Give incident-focused agents direct LogQL, PromQL, and Elasticsearch queries against your observability stack without building custom scripts each outage.
Overview
mcp-observability is a MCP server for the Operate phase that lets incident agents query Loki with LogQL, Prometheus with PromQL, and Elasticsearch from one stdio integration.
What is this MCP server?
- Query Grafana Loki with LogQL from MCP tools
- Run Prometheus PromQL for metrics-driven incident analysis
- Search and query Elasticsearch for log and event correlation
- Purpose-built for incident agents per marketplace description
- @infoinlet/mcp-observability v0.1.1 stdio npm package
- Version 0.1.1 npm package @infoinlet/mcp-observability
- Three query surfaces: Loki LogQL, Prometheus PromQL, Elasticsearch
- Repository path services/mcp-observability in infoinlet-marketplace
What problem does it solve?
When production breaks, you waste minutes opening Loki, Prometheus, and Kibana separately while the agent stays blind to real telemetry.
Who is it for?
Indie operators who already run Loki, Prometheus, or Elasticsearch and want Claude Code-style agents to investigate outages with native query languages.
Skip if: Greenfield projects with no observability stack, or teams that only use a single proprietary APM without these endpoints.
What do I get? / Deliverables
After MCP registration and backend credentials, the agent runs observability queries in-thread so you correlate logs and metrics faster during incidents.
- Unified MCP tools for LogQL, PromQL, and Elasticsearch investigation
- Agent-driven production triage without custom one-off query scripts
- Marketplace-packaged stdio server at version 0.1.1
Recommended MCP Servers
Journey fit
Operate is where logs, metrics, and search indices matter—this server is built for incident agents triaging production signals. Monitoring subphase covers Loki, Prometheus, and Elasticsearch query paths that replace manual dashboard hopping during fires.
How it compares
Multi-backend observability MCP connector, not a hosted monitoring product or SRE playbook skill.
Common Questions / FAQ
Who is mcp-observability for?
Solo builders and small teams on self-hosted or cloud Loki, Prometheus, and/or Elasticsearch who use AI agents for incident investigation.
When should I use mcp-observability?
Use it in operate and monitoring workflows when debugging production errors, capacity issues, or log patterns during or after deploys.
How do I add mcp-observability to my agent?
Install @infoinlet/mcp-observability (v0.1.1), configure stdio MCP in your agent, and set environment variables or config for each observability endpoint URL and auth your deployment requires.