Run three coding agents in real PTY terminals, turn tasks into agent sessions, resume native history, hand work across agents, and get notified only when a terminal needs attention. Local-first and MIT licensed.
Hi Product Hunt — I’m the maintainer of Termexo.
The first release focused on putting Claude Code and Codex into one recoverable Windows workspace. Version 0.6 turns that workspace into a more complete agent workflow:
- OpenCode is now a third first-class agent, with the same launch, resume, restart restoration, state tracking, and optional auto-confirm controls.
- Tasks can carry a project, agent, model, priority, and acceptance criteria, then launch as real PTY terminals and move from backlog to running, completed, and accepted.
- Prompt drafts and history are reusable assets, while sanitized handoff packages can carry Git context, verification, risks, and next steps between agents.
- Claude background sessions can be inspected, forked, or attached instead of failing on the first message after a normal resume.
- Tabs can be dragged, middle-clicked to close, switched with the wheel, and controlled from the keyboard.
Termexo remains local-first: no Termexo account, no cloud relay, native session files stay read-only, and API keys stay in Windows Credential Manager. It is Windows-only today and available under the MIT License.
I would especially value feedback from people who run several coding agents at once: where does your workflow lose the most context—task planning, approval, recovery, or handoff?
Replies
Termexo