1 · Choose the conversation
2 · The person opposite
What are you practising?
Pick a scenario. It sets who you're up against and the rubric you'll be scored on. Everything is editable below.
Grounds the scenario. The other party won't hand this to you.
3 · Connection
Anthropic key
The conversation and grading run against Claude directly from your browser. Stored only in this browser, sent only to api.anthropic.com.
No key saved yet
Voice feels best on a low-latency model.
Scoring rewards depth over speed.
4 · Voice & mic
How you'll talk
Hands-free by default: speak, pause, and it sends automatically. The other party replies out loud.
Browser voices vary by OS. For much better voices, pick a premium engine below.
★ = higher-quality neural/online voice, when your system has one.
Shorter = snappier but may cut you off mid-thought.
?
—
—
🎧
Ready when you are
Hands-free
speak & pause — it sends automatically
Trajectory
Score over time
No graded conversations yet.
By scenario
Where you stand
Run a graded call to populate this.
Development
Your development plan
Synthesised from every conversation you've logged.
Log at least one conversation, then generate a plan. The more you run, the sharper it gets.