What is Pytxo?

Pytxo coordinates coding agents on your machine. Describe a change once; Pytxo plans it into tasks, gives each task to an agent you chose, checks the combined result, and lets you review it before anything is applied. You keep your editor and your agents' accounts.

The technical name is an agent hypervisor: a control layer around the agent tools you already use.

What it adds

• A plan you can inspect: which agent runs each task, which files it owns, and which commands check the result.
• Separate workspaces: eligible runs prepare changes away from your working repository.
• A review step: inspect the prepared files and recorded checks before Apply.
• A recorded outcome: return to a run and see its execution and Apply status separately.

Tasks that claim the same file never run at the same time. Start with one agent, then put several agents on one job.

Choose how to use it

Desktop provides Work, History, Setup, and reviewed Apply on Windows x64.

CLI and MCP support terminal and editor workflows on Windows, macOS, and Linux. The CLI does not have an Apply command.

Local Core needs no Pytxo account. Agent sign-in, subscriptions, and API charges remain separate.

Know the limits

Apply controls repository integration for one root. It is not a guarantee that an agent cannot access the host or network. A run's enforcement receipt reports the protections that were active, advisory, unavailable, or bypassed.

Passing tests are evidence for those tests. An agent's own success message is a claim, and an interrupted Apply can require manual recovery.

Control over production APIs and other external effects is a proposed direction, not an available feature.

Start with Install, or learn how Review and Apply work.