Coldtea is a macOS desktop IDE that runs next to your repo and covers coding, testing, and production monitoring. You run CLI agents like Claude Code, Codex, or Gemini CLI in parallel git worktrees, and tasks sync with Linear, Jira, or GitHub Issues. Testing agents check your web and mobile apps on every PR. Monitoring agents turn production signals into tasks. It suits engineering teams who use coding agents and want to ship fast without regressions.

• Bring-Your-Own-Agent Terminal: Run any CLI agent, including Claude Code, Codex, Gemini CLI, and OpenCode, in a real terminal. You keep the agents and config you already use.
• Parallel Agents in Git Worktrees: Each agent gets its own git worktree, so several can work at once without conflicts. Agents can also message each other in split panes.
• Two-Way Task Sync: Tasks sync both ways with Linear, Jira, and GitHub Issues. Each task keeps its implementation plan and session logs, so you can see how the work was done.
• Built-In Editor, Browser, and Annotations: Review and edit code next to the terminal, and use an in-app browser that you and your agents can both control. Add notes to specific lines or panes and send them straight to an agent.
• Agentic QA Testing: You write tests in plain English, and a testing agent runs them on web, desktop, iOS, and Android. Tests can run on every pull request against the preview URL or on a schedule, and each run returns a verdict with evidence.
• Production Monitoring Agents: Agents read signals from Sentry, Datadog, Grafana, New Relic, PostHog, and Vercel. They investigate issues and file them as tasks for your coding agents.
• Optional Cloud Runs: Background agents can run tasks on a hosted machine, stream into a pane, and open a pull request. Cloud is off by default, so your code stays local until you turn it on.
• Shareable Artifacts and MCP Support: Agent runs can publish reports, previews, or docs as a link. Coldtea also works with MCP servers and your existing agent config.

• Free Core Product: The terminal is free forever with no seat limits or trial, and monitoring is free when you use your own provider or AI keys.
• Agent-Agnostic: You aren't tied to one model or vendor. Any CLI agent works, so you can switch between Claude Code, Codex, Gemini CLI, and others.
• Full development lifecycle in one app: Bring coding, code review, automated QA testing, and production monitoring together to reduce the need to switch between separate tools.
• Local-First by Default: Code stays on your machine unless you choose a cloud run, which helps teams with strict data policies.
• Plain-English automated testing: Describe test scenarios in natural language and let QA agents test web and mobile apps, with support for desktop testing as well. Review test results and supporting evidence.
• Easy access for teams: Get started without a sales call, seat minimum, or contact-only pricing tier.

• macOS Only: The desktop app is only offered for macOS, so developers on Windows or Linux can't use it locally.
• Brings No Agent of Its Own: Coldtea runs third-party CLI agents rather than shipping its own model. You still need access to and billing for an agent like Claude Code or Codex.
• Testing Is Credit-Based: Agentic testing includes 2,000 free credits per month, and heavier use on every PR across web and mobile means buying more credits.
• Monitoring Needs Your Own Keys: Free monitoring uses your own provider and AI keys, so you carry those usage costs and the setup.

• Parallel Feature Development: Give several tasks from Linear or Jira to separate agents, each in its own worktree. You review and annotate their changes in the built-in editor, so several features move forward at once.
• PR Regression Checks: Write QA tests in plain English and run them against every pull request's preview URL. Regressions get caught with evidence before merge, without a hand-maintained test suite.
• Mobile App QA: Have the testing agent drive your app on iOS and Android simulators. You can run bug bashes on a schedule and still release often.
• Production Error Triage: Connect Sentry or Datadog so monitoring agents investigate production issues and file them as tasks. Coding agents can then pick up fixes, which closes the loop from alert to pull request.
• Offloading Long-Running Work: Turn on cloud runs to send a task to a hosted machine. It streams progress into a pane and opens a pull request while your laptop stays free.
• Sharing Agent Output With Stakeholders: Publish an agent's report, preview, or doc as a link. PMs and reviewers can see the results without opening the IDE.