Introduction

Pytxo is a local-first agent hypervisor and telemetry plane. It coordinates headless terminal agents — Claude Code, Codex, Antigravity CLI, or any shell command — inside managed pseudo-terminals (PTYs), schedules dependency-aware waves, and logs structured telemetry to SQLite.

Pytxo is not an IDE and not a cloud virtual workspace. Your editor stays your editor. Pytxo plugs in through a local MCP hub and optional Reality Deck desktop telemetry.

What you get

• Parallel agents in isolated git worktrees without stomping the same files
• DAG wave scheduling so conflicting tasks run in separate waves
• Signal Core context scaffolding to cut noisy token spend (roadmap: tree-sitter AST skeletons)
• Blast Shield copy-on-write isolation and Race Shield stdin buffering (partial / rolling out per crate)
• MCP integration so Cursor or your IDE drives orchestration locally

Who this is for

Audience	Start here
New users	Install → First three-agent run
IDE users	MCP from Cursor
Operators	CLI reference, pytxo.toml
Contributors	Architecture, Contributing

Quick mental model

IDE / CLI  →  MCP hub  →  Orchestration (Rust)  →  PTY agents in worktrees
                              ↓
                    Reality Deck (optional telemetry UI)

Read What is Pytxo? for positioning vs cloud ADE products, or jump straight to Install.