Ava Supernova
AvaSupernova
In development· This page describes a mode we're actively building. It is not yet available in any release.
coming to the IDE — she'll stay on call

Your AI that actually shows up.

Error tracker fires at 2am. One tap on your phone. Ava opens your laptop, triages the bug, proposes the fix. 90 seconds, anywhere. Every other automation product stops when you go home.

This is what we're building. The foundations (persona orchestration, safety ontology, kill switches, real Playwright and OmniParser integrations) are already in. UI wiring and end-to-end trajectories come next.

persona teamsession whitelistuncacheable approvalstriple kill switchesrollback via git

Windows first. macOS next. Will be available on every tier, including Free, once it ships.

The difference

Everyone else has a clicker. We're building the brain that decides when to click.

👁️

Five specialists, one step

Scout observes. Planner decides. Actor executes. Verifier checks. Narrator tells you what just happened. Focused mindsets beat one model trying to do everything.

🧠

Memory across sessions

Ava will remember how to navigate your apps. Every other agent relearns them every time. Run a workflow twice and you'll feel the difference.

📱

Companion as remote

IDE stays alive in the tray. Your phone pairs over a secure channel. Send a command from anywhere. Approve irreversibles with a tap.

💻

Your machine, your credentials

Runs on your actual computer. Uses your existing logins. No cloud browser renting out sessions. Never uploads your screen content for training.

Safety by construction

Designed after reading every public autonomy incident. Every layer here exists because another product failed at it.

🛑

Irreversible actions always confirm — and approval never caches

Send, submit, pay, delete, publish. Each one will ask fresh every time. The industry caches these. That's the exact hole attackers exploit.

📋

Session whitelist, not persistent

At mode entry: "You may work in Gmail, VS Code, Azure." Any app outside the list pauses and asks. Doesn't remember between sessions. No "allow always" button.

Three kill switches

Triple-Escape global hotkey. Stop button always visible. Budget trip if a step/token/time cap hits. Any of them aborts the trajectory. All three will work from your phone too.

↩️

Git checkpoints before destructive ops

Where rollback is possible, it happens automatically. Where it isn't — emails sent, payments made — the confirmation prompt says so explicitly.

🔐

Secrets never hit the model

Passwords, API keys, 2FA codes handled via opaque handles. Ava sees "{secret}", not the value. Never in memory, trajectory log, or prompt context.

Cost you can predict

Target numbers from the spec — live counter during, hard caps default on. Final billing unit sets at launch once we pick an OmniParser host.

Typical task target
~200K
tokens · ~15¢
Hard cap default
500K
tokens / 30 steps / 5 min
Free tier will include
~15
medium tasks / month

The on-call flow we're building toward

How a 2am incident looks when Ava's in your pocket.

1
Sentry fires on your desktop. Error rate spike.
2
Companion push: "Production error rate spike — 127 errors in 5 mins."
3
Tap. Suggested prompt appears: "Find what's failing and propose a fix."
4
Tap Send. Companion pairs with your desktop. Trajectory runs. You watch on phone.
5
Approval card for the git commit: "Revert #4821?". Tap Approve.
6
Fix pushed. Deploy red → green. Ava summarises what she did.
7
You go back to sleep.

Where we are

Persona system: Scout, Planner, Actor, Verifier, Narrator — prototyped and token-measured
Safety ontology: five risk classes, uncacheable approvals, whitelist gates
Playwright subprocess lifecycle: validated with zero orphan processes on Windows
OmniParser integration: tested live on Microsoft's v2 via Replicate
Kill switches: triple-Escape global hotkey + tray mode + companion kill wired into the Rust layer
End-to-end trajectories: wiring the pieces together, exercising approval flow on real runs
Companion remote: admin-only private test, widening later
Public launch: when the kill switches have been exercised on real Windows bare-metal and the approval flow feels right

Want to be there when it ships?

Install the IDE today — it's already a full 61-tool coding agent. Desktop automation will unlock in the IDE for everyone, Free tier included, once we're ready to release it.