Work should start as an issue and end as a pull request.
We started Celiador because the gap between writing down what needs doing and having a reviewable change is still too wide. An agent can close it, if it works the way teams already do: an issue to brief it, a branch to work on, a pull request to review, and a person deciding what ships.
The first version of Celiador was a prompt-to-app builder. It hit the wall every prompt-to-code tool does: the moment you want to keep editing, the generated code drifts away from you. The second was an AI editor, a fork of VS Code with an agent in its core. That worked, but it kept us at the desk, watching a chat, for every change.
Today Celiador is the loop. You label a GitHub issue celiador; an agent runs it on your Mac in its own git worktree, comments on the issue, and opens a pull request that closes it; you review and merge from your phone or from the editor's Agents view. The editor is still here, as the nicest runner. VS Code and Cursor run the same loop through the Celiador Remote extension. Cloud runs for when your Mac is off are being built. It's the tool we wanted for ourselves, and every change to Celiador now ships through it.
You merge, the agent proposes
The agent works on a branch of its own and the result is a pull request. Nothing lands until a person reads it and merges. Speed shouldn't cost you trust.
Keep your editor
The Celiador editor is the nicest runner, but the loop also works from VS Code and Cursor through the Celiador Remote extension. We would rather meet you where you are than ask you to move.
GitHub is the record
The run narrates itself on the issue and the pull request closes it. Your team sees what the agent did where they already look, not in a side tool.
Your code is yours
Runs happen on your machine. We don't train on your code. Privacy is a default, not an upsell.
Come build with us.
We're a small team. Try the loop on one issue, or say hello.