Coldtea
Coldtea is an agentic development environment for macOS that unifies terminal, end-to-end testing, and production monitoring into one self-driving software delivery platform.
At a Glance
For solo builders and side projects. Terminal, tasks, and coding agents are free forever.
Engagement
Available On
Alternatives
Listed Aug 2026
About Coldtea
Coldtea is a desktop IDE for macOS that brings together terminal-based coding agents, visual QA testing agents, and production monitoring agents into a single agentic development environment. It targets the full software delivery loop — not just code writing — by covering regression testing, deployment gating, and production incident detection. Currently available for macOS, with Windows and Linux on a waitlist.
What It Is
Coldtea positions itself as an "agentic development environment" that goes beyond code generation. Where most AI coding tools focus on writing code faster, Coldtea covers what happens after: automated QA on every pull request, production monitoring that investigates regressions and opens pull requests with fixes, and a terminal that lets multiple coding agents collaborate as a team. The product runs locally on your machine, with optional cloud execution for background agent tasks.
How the Delivery Loop Works
Coldtea organizes work into four layers:
- Terminal: Run multiple coding agents (Claude Code, Codex, OpenCode, and others) side by side. Agents can plan, build, review, and improve work together while the developer stays in control.
- Agentic Testing: Describe web or mobile user journeys in plain English. QA agents convert them into self-healing tests, run the full suite against every PR preview URL, and gate deployments on the result. Web runs cost 10 credits; mobile runs (using simulators with real builds) cost 100 credits.
- Self-Driving Monitoring: Always-on monitoring agents watch production across error logs, customer feedback, and user sessions. When they find something, they investigate and open a pull request for review. They also scan the codebase for technical debt introduced by agent-written code.
- Tasks: Sync engineering boards (Linear, Jira, and others), assign tasks to background cloud agents, and track progress without leaving the app.
Architecture and Agent Compatibility
Coldtea is designed to work with agents the developer already uses rather than replacing them. It resolves the terminal shell automatically and carries over existing settings. Code stays on the developer's machine by default; cloud execution is opt-in. The platform supports Claude Code, Codex, OpenCode, and any other CLI-based agent. Monitoring integrations include Sentry, Datadog, Grafana, New Relic, PostHog, and Vercel, with new providers typically added within 24 hours of a request.
Platform and Availability
Coldtea is currently available as a macOS desktop application. Windows and Linux support is in progress via a waitlist. The app connects to GitHub for pull request workflows, and to Linear and Jira for task board sync. An MCP tool catalog is documented in the reference section, indicating MCP server integration for agent tooling.
Why It Matters
The core argument Coldtea makes is that the bottleneck in AI-assisted development has shifted from writing code to verifying and operating it. Coding agents now produce code faster than QA and monitoring tooling can keep up with, and Coldtea's homepage states that "speed stopped being the problem — the risk moved downstream, into regressions nobody caught and incidents your users report first." By putting build, test, and monitor in one environment, Coldtea aims to let teams ship continuously without fragmenting across separate tools for each phase of delivery.
Community Discussions
Be the first to start a conversation about Coldtea
Share your experience with Coldtea, ask questions, or help others learn from your insights.
Pricing
Free
For solo builders and side projects. Terminal, tasks, and coding agents are free forever.
- Unlimited tasks, synced from Linear, Jira and more
- Works with any CLI agent: Claude Code, Codex, OpenCode
- Cloud runs for background agents
- 2,000 agentic credits a month (~200 test runs)
- 2 connected tools per project
Pro
For teams shipping every day. First seat is free, each additional teammate is $20/month.
- Everything in Free
- First seat free, every teammate after is $20/month
- Shared boards, assignment and review across the team
- Add Agentic testing Pro for 10,000 credits a month
- Add self-driving per project from $49/month
- Unlimited connected tools on any project running self-driving
- Top up credits any time, no hard cap
- Cancel from inside the app
Enterprise
For organizations with security and compliance needs.
- Everything in Pro
- Volume pricing on credits and self-driving projects
- Annual billing and invoicing
- Priority support and hands-on onboarding
- Shared Slack channel
Agentic Testing Pro
10,000 credits a month (~1,000 test runs). Replaces the free allowance of 2,000 credits.
- 10,000 credits a month (~1,000 test runs)
- Replaces the free allowance of 2,000 credits
- Add 10,000 more credits any time for $120
- Purchased credits roll over and do not expire at renewal
- 10 credits per web run, 100 per mobile run
Self-driving
Monitoring agents watch production, investigate findings, and open pull requests. From $49/project/month.
- Essentials ($49/month): watches signals and raises what matters most
- Complete ($99/month): deeper reasoning across signals and code
- Lifts the 2 tool cap on the project it is turned on for
- Every project starts on a 7 day trial
Capabilities
Key Features
- Multi-agent terminal with team coordination
- Agentic QA testing for web and mobile
- Self-healing tests from plain English descriptions
- PR-gated automated regression testing
- Production monitoring agents with auto-investigation
- Automatic pull request creation for fixes
- Technical debt detection in agent-written code
- Cloud runs for background agent tasks
- Task board sync with Linear and Jira
- Works with Claude Code, Codex, OpenCode, and any CLI agent
- Monitoring integrations: Sentry, Datadog, Grafana, New Relic, PostHog, Vercel
- MCP tool catalog support
- Local-first with optional cloud execution
Integrations
Demo Video

