Mel – the agentic terminal
The AI terminal for teams.
Security & privacy
Trusted with your code.
Mel runs its agent's tools on your computer, keeps provider keys out of the app on paid plans, stores conversations on your machine and in your account if you keep sync on, and lets you choose which steps need your approval. Here's exactly how Mel handles keys, code, and access.
How it works
Transparent by design.
Where keys live
Hosted-plan provider keys are held server-side on the Mel relay — the app ships with none. If you bring your own key, it is stored on your device — in the OS keychain, or a file only you can read on Linux — and travels with each request to the relay, which uses it for that request and never stores or logs it.
What leaves your machine
Agent tools — read, edit, run — execute locally. Only the context a given model call needs is sent to the relay, which calls the provider.
Where your data lives
Conversations and history are stored on your machine, under your home directory. A copy of each saved conversation also syncs to your Mel account unless you turn that off in Settings → Privacy.
Configurable approvals
Set in Settings > Agent mode what the agent may do without you. A command Mel recognises as impossible to undo, a command or file read that touches credentials, a write outside your project and a connector call stop for you whatever you choose — and since v0.1.63 Mel follows each cd on a command line, so a delete in another folder is caught too.
Signed updates
Every update must be signed by Mel; one that isn't is refused and you keep the version you have. From v0.1.63, stable updates are signed with a new key.
Open source: planned
The terminal and relay are built to be readable and auditable. Open source is planned, not available yet.
Report a vulnerability
Found a security problem? Tell us privately.
Email security@openmel.dev and put SECURITY at the start of the subject line. Please do not open a public issue or post the details before the problem is fixed.
What helps us
- What you found, and where: the part of Mel, the page or the endpoint.
- How to reproduce it: your Mel version (Settings > About), your operating system, and the steps.
- What someone could do with it, and whether anyone else knows about it yet.
What to leave out
- Other people's data. Send only what is needed to reproduce the problem.
- Real keys, tokens and passwords. If one is involved, revoke or rotate it first, then tell us that it existed and where you found it.
- Anything from accounts you do not own. Test only with your own account and data.
This is also published in machine-readable form at /.well-known/security.txt.
Get started
Built to be trusted with your code.
The free Mel terminal runs local-first today. The AI agent is free with your own API key. Hosted keys stay server-side, your own keys are stored on your device and never on Mel's servers, and you choose which steps need your approval.