Coldtea

Coldtea is a desktop workspace for teams that already ship with coding agents and want the rest of the delivery loop — testing, release gating, production monitoring and repair — handled by agents too. It keeps your existing shell, dotfiles and model of choice, and wraps them in a UI where several agents can plan, build, review and critique each other's work while you stay the decision-maker.

The product is organised around four surfaces. Terminal runs multiple coding agents side by side — Claude, OpenAI Codex, OpenCode and others — with a reviewer agent that can block a change before it lands. Agentic Testing turns a plain-English description of a user journey into a browser test that repairs itself when the UI moves, then runs the suite against PR preview deployments so a broken flow gates the merge instead of surfacing in production. Production monitoring watches error logs, session replays and customer feedback for regressions, investigates what changed and opens a pull request with a proposed fix. Tasks syncs with your engineering board and dispatches background cloud agents to work items in parallel.

It reads from the observability stack you already run — Sentry, Datadog, Grafana, New Relic, PostHog and Vercel are all supported sources — so the monitoring agents reason over real signals rather than a parallel telemetry pipeline you would have to set up first.

Privacy model: Coldtea runs locally by default and your code is not moved to the cloud unless you explicitly turn on cloud runs, which matters if your repository cannot leave your machine.

Pricing: the terminal is free permanently, agentic testing has a free tier, and production monitoring is free to trial. Coldtea says its users include engineers at Revolut, HelloFresh, Amazon, SuperTokens and Appsmith.

Who it is for: engineering teams that have adopted agentic coding and are now bottlenecked on everything downstream of the diff — regression testing, triage and the small production fixes nobody has time to write.