You set the bar. Codus holds it.
Sliders for testing rigor, security review and response detail — plus a 2D style pad for visual fidelity. Global defaults, per-project overrides, applied live.
Here is what happens, step by step.
- 01step
Set the defaults once
Three sliders in Settings — testing, security, response detail — plus the style pad. This is the bar every agent builds to by default.
- 02step
Override per project
A prototype can run loose while production stays strict. Overrides attach to the folder, so they follow the project into any quadrant.
- 03step
Resolved at verification time
The settings are read fresh every time an agent verifies its work, so dialling rigor mid-session changes the very next check. No restart, no re-prompt.
"AI ships broken code" is a settings problem, not a model problem.
Most tools hard-code one verification behaviour for every project — too sloppy for production, too pedantic for a spike. Codus makes the quality bar a dial you own, and the agent's definition of done follows it.
More of what makes codus different.
Queue a feature. Walk away. Get screenshots.
Assign a task from your phone. Codus runs it on your Mac. Screenshots in three viewports land in your notifications when it's done.
Captchas, 2FA, passkeys. All solved.
Cloudflare Turnstile, hCaptcha, reCAPTCHA. All auto-solved. TOTP codes auto-typed. WebAuthn passkeys auto-stored. Agents log in everywhere, every time.
Run 6 dev servers at once. Forget port 3000.
Per-folder port registry. Auto-rebinds your Valet and Herd setup when ports drift. The "port already in use" error is dead.
Try codus on your codebase.
£99 per seat / month. Unlimited usage. Bring your own Claude Code or Codex account.