PytxoIntroduction
Pytxo is a local agent hypervisor for the coding agents you already use.
Pytxo is a local agent hypervisor: it runs the coding agents you already use (Claude Code, Codex, Antigravity CLI, or any shell command) in the background, schedules them so they do not overwrite each other, shows what they changed, and lets you approve merges before those changes land in your real repo.
Pytxo is not an IDE and not a cloud workspace. Keep your editor. Connect through a local MCP hub, or use the optional Pytxo Desktop app.
At a glance
- Run several agents at once in isolated copies of your repo
- Schedule conflicting tasks in separate waves so they never collide
- Send agents structural code skeletons instead of whole files (fewer tokens)
- Approve risky commands and merges yourself (Galaxy tier)
- Workspaces: one run across API, web, and shared protos
- Fleet runs: coordinate work across separate repos when you need hard barriers
- Drive it from Cursor or another IDE through local MCP
Who this is for
| You want to... | Start here |
|---|---|
| Install and run your first swarm | Install -> First three-agent run |
| Use Cursor or another IDE | MCP from Cursor |
| Operate from the terminal | Hypervisor Shell |
| Look up commands and config | CLI reference, pytxo.toml |
| Fix doctor / trust / MCP issues | Troubleshooting |
Hypervisor Shell
Run pytxo with no subcommand to open the Hypervisor Shell. Trust a folder, preview waves, start agents with your own API keys, and approve risky actions. See Folder trust and Hypervisor Shell.
Quick mental model
IDE / CLI -> MCP hub -> Orchestration -> Agents in isolated copies
↓
Pytxo Desktop (optional control UI)Read What is Pytxo? for the full picture, or jump to Install.