chmonitor
Star Open Cloud
Open source · GPL-3.0

A simple, beautiful UI to monitor ClickHouse

chmonitor turns your cluster's system tables into a fast, readable dashboard — queries, merges, parts, replication and health, plus an AI agent that answers questions about your data. Self-host it, or use the hosted cloud.

Open the cloud dashboard Star on GitHub
Overview
chmonitor cluster overview
Cluster Topology
chmonitor cluster topology
Health Summary
chmonitor health summary
AI Agent
chmonitor AI agent
Cluster Insights
chmonitor cluster insights
Running Queries
chmonitor running queries
Explain Query
chmonitor explain query
Most Expensive Queries
chmonitor most expensive queries
Data Explorer
chmonitor data explorer
Metrics
chmonitor metrics
What you get

Everything about your cluster, in one place

chmonitor reads ClickHouse system tables directly — no exporters, no extra storage. Each view is purpose-built so you can spot the problem and move on.

Ask your cluster anything

The built-in AI agent connects over MCP and answers questions about schemas, storage, query performance and health — then links you straight to the matching view.

  • Chat about schema, storage, slow queries and system health
  • Bring your own model, skills and MCP tools
  • Read-only MCP endpoint for Claude, Cursor or any client
chmonitor AI agent chat
chmonitor MCP server setup

Every query, explained

Running, history, failed, slowest and most expensive queries — each with duration, memory, rows read and the full statement. Filter by user, slice by time and find the culprit fast.

  • Queries by user and top query patterns over time
  • Saved filter presets and one-click EXPLAIN
  • Kill long-running queries right from the table
chmonitor running queries with live charts and table

See the whole cluster at a glance

An interactive topology map of nodes, shards, replicas and the Keeper quorum — with live replication and coordination links, health states and an inspector for every cluster.

  • ClickHouse nodes and Keeper quorum with leader & followers
  • Physical and logical clusters, overlapping virtual clusters
  • Per-node CPU, memory and latency, healthy / warn / unreachable
chmonitor cluster topology map
Data Explorer

Explore tables and how they connect

Browse every database and table, map dependencies between them, then drop into a SQL editor to query the data — all without leaving the dashboard.

chmonitor data explorer dependency graph
01
Dependency graphSee materialized views, dictionaries, joins and sources wired between tables — vertical or horizontal.
chmonitor data explorer SQL query editor
02
Query editorRun SQL with row counts and timing, inspect structure, DDL, indexes and parts for any table.
Health Summary

Catch issues early — then brief your agent to fix them

Real-time health indicators for the whole cluster — replication lag, failed queries, disk and memory, stuck mutations and more. When something turns red, generate a ready-to-paste prompt that briefs your AI agent to debug it.

Every health signal, color-coded

A single board of indicators — green, warn or critical at a glance — each linking straight to the views that explain it.

  • Replication lag, readonly replicas and delayed inserts
  • Failed and long-running queries, OOM kills, Keeper exceptions
  • Disk, memory, parts-per-partition and mutation status
chmonitor health summary indicators

A ready audit prompt for your agent

One click turns any failing check into a structured prompt — the metric query, raw data, relevant system tables, common causes and docs. Paste it into Claude, ChatGPT or your coding agent for a tailored diagnosis and remediation plan.

  • Toggle exactly what context to include
  • Live values, thresholds, host and version baked in
  • Copy as Markdown — works with any AI tool
chmonitor generated audit prompt dialog
And a lot more

One dashboard for the whole cluster

From a high-level overview down to ZooKeeper logs and replication mirrors — every system table you reach for, already laid out.

Overview & health

CPU, memory, connections and replication status at a glance, with health checks across the cluster.

Cluster insights

Record-breaking scans, busiest days, total rows read and storage — the highlights from all your data.

PeerDB replication

Watch mirrors and peers with live status, rows-synced trend and routing — failures surface immediately.

Merges & parts

A live part-lifecycle timeline with merges, mutations, moves and part-size distribution by table.

Security & access

Users, roles, grants and quotas — review who can touch what, all from the same place.

Logs & system

Query, error, trace and ZooKeeper logs alongside settings and metrics — searchable and time-ranged.

Open source

Self-host it, or use the cloud

chmonitor is open source. Deploy it to Cloudflare Workers, run it as a Docker container, or install the Helm chart on Kubernetes — or skip setup entirely and connect through the hosted cloud. Same UI, your choice.

  • Up in minutesPoint it at a ClickHouse connection string and you're monitoring.
  • Read-only by designConnect with a least-privilege user — chmonitor only reads system tables.
  • Yours to extendGPL-3.0 licensed, community-driven, contributions welcome.
View on GitHub Read the docs

Edge hosting via the OpenNext adapter — no servers to run.

bash
$ git clone https://github.com/duyet/clickhouse-monitoring.git
$ cd clickhouse-monitoring
$ bun install
 
# production env + a token with Workers deploy permission
$ cat <<'EOF' > .env.prod
CLOUDFLARE_API_TOKEN=<token>
CLICKHOUSE_HOST=https://clickhouse.example.com:8443
CLICKHOUSE_USER=default
CLICKHOUSE_PASSWORD=change-me
EOF
 
$ bun run cf:config   # push secrets to the Worker
$ bun run cf:deploy   # build + deploy + populate cache

Single container from GHCR. Pin a release tag in production.

bash
$ docker run -d --name chmonitor -p 3000:3000 \
  -e CLICKHOUSE_HOST='https://clickhouse.example.com:8443' \
  -e CLICKHOUSE_USER='default' \
  -e CLICKHOUSE_PASSWORD='change-me' \
  ghcr.io/duyet/chmonitor:latest
# open http://localhost:3000
Trial with a bundled ClickHouse
bash
$ git clone https://github.com/duyet/clickhouse-monitoring.git
$ cd clickhouse-monitoring
$ docker compose up -d
ClickHouse running on the Docker host? localhost points at the container — use host.docker.internal (--add-host=host.docker.internal:host-gateway on Linux).

Chart vendored at deploy/helm/chmonitor — Secret-backed password, non-root uid 1001, health probes.

bash
$ git clone https://github.com/duyet/clickhouse-monitoring.git
$ cd clickhouse-monitoring
 
$ helm install my-chm ./deploy/helm/chmonitor \
  --set clickhouse.host="https://clickhouse.example.com:8443" \
  --set clickhouse.user="default" \
  --set clickhouse.password="change-me"
 
$ kubectl port-forward svc/my-chm-chmonitor 3000:3000  # <release>-chmonitor
Upgrade / remove
bash
$ helm upgrade my-chm ./deploy/helm/chmonitor -f my-values.yaml
$ helm uninstall my-chm
For production, keep the password out of --set (use clickhouse.existingSecret), enable Ingress, and set autoscaling.enabled=true.

Start monitoring in minutes

Open the hosted dashboard and connect a cluster, or grab the source and run it yourself.

Open the cloud dashboard View on GitHub