Privacy policy
What we touch. What stays yours.
Your code and files stay on your computer. When you use AZMX Cloud (AZMX Auto), the messages you send to the AI pass through AZMX's gateway to our cloud AI provider to get an answer; we don't sell them or train on them. With your own key, requests go to that provider. With Local-Only mode on, nothing leaves your computer. No account needed.
What we touch
The very short list.
License checks
Only if you activate a licence: renewal + revocation check with the licence token and device id. No prompts, no code.
Update manifest
One of the two automatic requests: a public, signed release manifest on GitHub, at most every 30 minutes. Blockable in air-gap.
Sync ciphertext
Pro+ opt-in. We see ciphertext + license id only.
AZMX Cloud
On by default. AZMX Auto, the default model, answers through AZMX's gateway and our cloud AI provider, and coding on it is free. The device gets an anonymous token wallet (no account). Local-Only mode turns it off.
What we don't
What we don't do with your data.
Source code
Your files stay on your computer. Only what goes into a message is sent: through AZMX's gateway when you use AZMX Auto, straight to the provider when you use your own key, nowhere with Local-Only mode.
Prompts
With AZMX Auto they pass through our gateway to get an answer — not sold, not trained on. With your own key they go straight to that provider.
API keys
0600 file. Never keychain. Never logged. Never synced.
Audit log
Local. Pro+ exports on demand — never auto-syncs.
Files read
Read locally. Secret-screen blocks .env/.ssh/credentials.
Selling & training
We don't sell what you send and we don't train on it.
Every request, listed
What the app sends, and when.
Automatic (default settings): the signed update check (GitHub release manifest), and registering this device's anonymous AZMX Cloud wallet (a device id, no account). Those are the only requests the app makes without you doing something.
When you send a turn: your prompt, attached context and learned preferences go to the AI you chose: through AZMX's gateway to our cloud AI provider for AZMX Auto (the default), or straight to the provider when you use your own key. Editor autocomplete does the same when you have a key for its provider. The model list of each provider you added a key for is refreshed about twice a day.
When you use a feature: web search and page fetch (when the agent uses them in your turn), research crawling (when you start it), MCP connectors (only servers you added), the connector directory (when you open Settings → Connectors), the on-device model download (when you click Download), and installing a missing language server when you open a file (can be turned off).
Only if you turn it on: cloud fallback for local features (embeddings, rerank, vision, voice, workspace indexing), Pro sync (end-to-end encrypted), and licence checks (after you activate a licence). AZMX Cloud itself is on by default so that free coding with AZMX Auto works without setup.
Local-Only mode is the one privacy switch: it blocks every cloud AI provider, AZMX Cloud and every optional network feature, leaving only the signed update check. With it on, none of your code, files or messages leave your computer.
On-device training data is off by default: nothing from your chats or dictation is saved for training unless you enable it, and it never leaves your machine.