Lopy
The control room for your coding agents.
A desktop app and a local engine that drive coding-agent CLIs through a real delivery lifecycle. Plan the work, run several agents at once in isolated git worktrees, watch each session stream, review every changed file, then merge.
- Who it is for
- Developers and small teams running AI coding agents daily.
- Stage
- Beta
- We build everything else with it, every day.
- Built with
- ElectronReactTypeScriptPythonFastAPIGit
The problem
Why this product exists.
A coding agent in a terminal gives you one session, no history, and no review step before it touches your branch.
Running three at once means three sets of changes fighting over one working tree.
When an agent needs a decision halfway through, there is nowhere for it to ask, so it guesses.
What it does
The engine makes zero LLM calls
Lopy's engine builds command lines, parses output, runs deterministic gates and stores artifacts. The intelligence stays in the agent CLIs you already pay for, so there is no second model bill and no second place for things to go wrong.
Session-owned worktrees
Each task owns a git worktree for its whole life — reset when it restarts, reused on a follow-up, merged only when you say the task is done.
Review before it lands
A Changes panel with a multi-file diff viewer and one honest action per session: merge to main, which only marks the task done if the merge actually landed.
Board, plan and run
Put work on a board, approve the plan, then run it. Each task opens its own visible streaming session, with a command palette and keyboard shortcuts throughout.
Your agents, your rules
Manage agents, skills, instructions, hooks and MCP servers in one place, and pick the agent, model and permission mode per session.
Code lives in git only
The store holds metadata and git references. It never holds your source, so removing Lopy leaves your repository exactly as it was.
Most tools wrap a chat box around an agent. Lopy models what actually breaks once you run more than one: isolation, scheduling, review and an honest definition of done.
Interested in Lopy?
Tell us about your setup and we will show you it running on data like yours — not on a demo dataset.