chmonitor/chmonitor

Open-source operational advisor for ClickHouse — real-time monitoring plus AI-driven index/partition/materialized-view recommendations.

chmonitor – Operational Advisor for ClickHouse

What it is – A web‑based dashboard (and a terminal UI) that connects to one or more ClickHouse servers, reads the built‑in system.* tables and presents a live view of cluster health, query activity, storage usage, replication status, etc. It also includes an AI‑powered “advisor” that can suggest schema or query optimisations such as projections, skip indexes, PREWHERE clauses or materialised views.

Key capabilities

Feature What you get
Cluster overview Topology map, health checks, >30 real‑time charts showing query volume, latency, data written, merges, etc.
Query insight Lists running, slow, failed and historic queries; read‑only SQL console with history and EXPLAIN.
Storage & data Explorer for table sizes, parts, compression ratios, TTL settings.
Replication health Views of merges, mutations, replication queues and ClickHouse Keeper quorum.
AI advisor Built‑in chat‑style agent (Claude / Cursor via MCP) that can answer English questions about schema, performance and suggest optimisations.
CLI / TUI chm command provides a terminal UI and a chm doctor health‑check shortcut.
Multi‑host support Monitor several ClickHouse instances from a single dashboard.

How to run

  • Docker (quickest): docker run -d -p 3000:3000 -e CLICKHOUSE_HOST=… -e CLICKHOUSE_USER=default -e CLICKHOUSE_PASSWORD=change-me ghcr.io/chmonitor/chmonitor:0.3.7 then open http://localhost:3000.
  • Other deployments: Helm chart for Kubernetes, Cloudflare Workers, or one‑click scripts for Railway/Render/Fly. Local development instructions are in the docs.
  • CLI: curl -sSf https://chmonitor.dev/install.sh | bash && chm doctor gives a terminal view.

AI integration The dashboard ships an “AI agent” that talks to Claude or Cursor through the MCP protocol. Users can type natural‑language questions (e.g., “How can I speed up this query?”) and receive concrete suggestions (add a projection, create a skip index, etc.). The agent runs locally; it does not execute DDL automatically – recommendations are advisory only.

Licensing & editions

  • Core software is GPL‑3.0 and can be self‑hosted for free with unlimited hosts.
  • Optional commercial license provides a support contract and the ability to list the buyer as a sponsor; it does not unlock extra features.
  • Pricing tiers: Personal (free), Team ($499/yr), Unlimited ($999/yr) with corresponding lifetime purchase options.

Who it’s for

  • ClickHouse DBAs and developers who need a single pane‑of‑glass view of cluster performance.
  • Teams that want AI‑driven hints without giving the tool write access to the database.
  • Anyone running ClickHouse on‑prem, in Docker/K8s, or on ClickHouse Cloud.

Where to learn more


All information above is taken directly from the repository’s README.

Related

  • Project
  • Project
  • Project
  • Project