
ProbeDeck: ClickHouse® Client
Cluster Monitor, Kill Query
0 ratings
Free
About
ProbeDeck is a secure mobile client compatible with ClickHouse® servers, built for cluster operations from your iPhone.
ProbeDeck is built for DevOps, SRE, platform and data engineers on call. Connect over TLS, through an SSH tunnel or a bastion. See cluster health, find the query that is eating the cluster and kill it in seconds, while production credentials never touch someone else's server.
WHY PROBEDECK
• Trust-first. Your credentials stay on your device and never leave it. There is no backend, no proxy, no traffic relay. Your connection goes straight from your phone to your server.
• It knows ClickHouse. Replication lag, read-only replicas and Keeper trouble, parts and merges, "too many parts", stuck mutations, disk pressure. These are the pages you get, so this is the first screen you see.
• Safety on each statement. ProbeDeck classifies your SQL before it sends anything. DROP, TRUNCATE, ALTER UPDATE and ALTER DELETE, DETACH, DROP PARTITION, OPTIMIZE FINAL and KILL need a type-to-confirm step showing the exact target. Costly statements carry a warning. Set a connection read-only and unlock writes when you mean to.
• Honest about what cannot be undone. KILL MUTATION stops a mutation but does not roll back what it already wrote. ProbeDeck says so instead of offering a fake Undo.
• Built for mobile networks. Reconnect with backoff, path-aware monitoring of Wi-Fi and cellular handoffs, and specific errors in place of an endless spinner. A ClickHouse error in the response body reads as a query error, not as a dead network.
• A production badge marks each risky action, so you know which cluster you are about to touch.
WHAT YOU CAN DO
• Connect over TLS, an SSH tunnel or a bastion host, with a Face ID app lock.
• Inspect a self-signed or custom-CA certificate fingerprint and accept it per connection.
• Read cluster health at a glance: replicas, disks, parts, merges, mutations and metrics.
• List running queries with elapsed time, memory and the user behind each one.
• KILL QUERY on the query burning the cluster, KILL MUTATION on the one that is stuck.
• Browse databases, tables, columns and DDL.
• Run SQL in the editor and read results laid out for a phone screen.
• Ask the AI assistant why a replica is lagging, through your own Ollama or your own API key.
PRIVACY
• Local-first. Your hosts, passwords, SSH keys and certificates stay on your device, in the iOS Keychain.
• No secrets in logs. ProbeDeck keeps passwords, hostnames, database, table and column names and values out of its logs.
• The AI assistant talks to your Ollama or your key. Nothing reaches us, because we run no server. Schema context is filtered by an allow-list before it goes to a model, and the assistant never executes anything by itself.
• Crash and error reporting is on by default and can be turned off in Settings. Reports are scrubbed and never include hostnames, table names, values or credentials.
• No tracking, no IDFA and no advertising identifiers.
ProbeDeck is foreground-first. It fits on-call inspection and operations rather than always-on background alerting.
FREE VS PRO
ProbeDeck is free to try, with a real, usable free tier, not a locked demo.
• Free: connect over TLS, SSH tunnel or bastion; the whole monitoring surface, including replicas, disks, parts, merges, mutations and metrics; browse schema and data; run read SQL; one saved connection.
• Pro unlocks operating: KILL QUERY and KILL MUTATION, write and DDL SQL, the AI assistant, and multiple saved connections.
"ProbeDeck Pro" is a one-time unlock. No subscription, no ads, no account required.
ProbeDeck works with self-hosted ClickHouse, ClickHouse Cloud and clusters behind an SSH bastion, over the HTTP interface on ports 8123 and 8443.
ClickHouse is a registered trademark of ClickHouse, Inc. ProbeDeck is not affiliated with, endorsed by, or sponsored by ClickHouse, Inc.
Show more
+2
What's New in ProbeDeck
1.1.1
August 3, 2026
Long queries work now. • A running query shows rows read and percent complete. No more silent spinner. • Anything slower than about 12 seconds used to fail with "Server unreachable" on a perfectly healthy server. Fixed. • You get 60 seconds instead of 30. Leave the screen and the server stops your query, so it no longer burns cluster CPU behind your back. • Read-only ClickHouse accounts work everywhere in the app. Before, they connected but could not run a single query. • When the server cuts a query short, you see the error. Earlier builds showed the partial rows and said nothing. • The app tells a lost connection apart from a query the server refused.
More







