Security
Your keys never leave your computer.
Most monitoring tools ask you to hand over your accounts. Periskopos runs on your PC and talks to each provider directly: no server of ours in between, and no account to sign up for.
How it's built
Nothing between you and your providers.
Each key is read from one file in your user profile and sent only to the provider it belongs to. Nothing goes through a server of ours, because there isn't one.
Read the threat model →- Listens only on your machine
The server binds to 127.0.0.1 and opens only in its own window, with a key that exists in memory for one launch.
- Keys go only to their provider
No page, tool or backup ever returns one.
- Read-only by default
Actions are opted into per account and always confirmed. Dangerous ones ask you to type the app's name.
- Open source
Every line is on GitHub under Apache-2.0, so you can check all of this yourself.
AI assistants
Your AI assistant can check. Only you can act.
Ask Claude, Cursor or VS Code "why is CI failing?" right where you code. They connect through MCP, never see a key, and can only ask to run an action: you approve each one in Periskopos. Limit any assistant to some apps, or to reading only.
you is my last push to Orbit-API live?
get_app("orbit-api")
assistant Not yet. e81b6d2 was pushed 3 minutes ago and CI passed, but the deploy failed on Render: npm ci couldn't resolve a package. Want me to redeploy?
run_action("orbit-api", "redeploy")
Waiting for your approval in Periskopos
FAQ
Keys and assistants.
The whole picture is in the threat model. Found a hole? Report it privately.
Where are my keys stored?
In one file in your user profile, outside any project: never in the database, a backup or the browser. Each key is only ever sent to the provider it belongs to. More on keys.
What access do the keys need?
One key per account, read-only wherever the provider allows it. Providers lists every account it can read and the exact permissions each key needs.
Can my AI assistant break something?
It can only ask. Assistants read what Periskopos sees and can request an action, but you approve each one in Periskopos, and actions are off for every account until you switch them on. Read the threat model.
Keep watch all around.
Free for as many apps as you run. Add your first one from its address and see its uptime, certificate and configuration in under a minute.
v0.5.0 · What's newThe first time you open it
- Windows
- SmartScreen says "unknown publisher": choose More info, then Run anyway.
- macOS
- It won't open at first: choose Done, then Open Anyway in System Settings → Privacy & Security.
- Linux
- Make the AppImage executable with
chmod +x, then run it.
These go once the app is code-signed (code signing policy).
Watchtower: an optional hosted watcher that keeps checking while your computer is off, and wakes you when it matters. Planned at $6 a month. Join the waitlist →