Issue in,
pull request out.
Label a GitHub issue celiador. An agent runs it on your Mac in its own git worktree, opens a pull request that closes the issue, and you review and merge from your phone.
Free in early access · Claude, on our key · Limits published
Keeping VS Code or Cursor? Add Celiador Remote and you get the loop.
Use the existing ApplePay helper in pay/. Do not touch the cart.
Added an ApplePay.authorize call in checkout.ts and wired the token into capture(). Lint and tests pass locally.
What happens when you add the label
An agent, a worktree, a pull request, and a phone to review it from. The editor is one way to run it, not the point.
The loop
Label a GitHub issue celiador. An agent picks it up on your Mac, works in its own git worktree so your checkout is untouched, and opens a pull request that closes the issue.
Review from your phone
Celiador for iPhone shows what the agent did, the diff, and what needs you. Send a follow-up, allow or deny a risky command, then approve and merge. The editor's Agents view shows the same runs.
Your runner, your choice
Run it in the Celiador editor, or keep VS Code or Cursor and install the Celiador Remote extension. Same loop, same account, same phone.
comingCloud runs for when your Mac is off. They need the Celiador GitHub App on the repository.
GitHub is the cockpit
Every run narrates itself on the issue: a comment when it is picked up and on which branch, and the pull request link when it is done. Your team sees the work where they already look.
comingThe agent's plan and its questions posted on the issue, answered by replying there.
Every PR explains itself
The pull request says what the agent did and closes the issue it came from, so review starts from the agent's own account of the change.
comingA self-review section in every PR, per-run cost on the PR and on the card, and failed CI checks fixed by the agent.
Our key, your privacy
No client holds a model key. Requests go through our server, which holds the key and meters usage with limits we publish. Your code isn't training data.
Three steps. Two of them are yours.
You write the issue and you merge the pull request. The agent does the part in between, on a branch of its own.
Label it
Write the issue the way you would brief a colleague, then add the celiador label. Within a minute the issue gets a comment saying it was picked up and on which branch.
It runs in its own worktree
The agent reads, edits, runs your lint and tests, and commits on a fresh branch in a git worktree on your Mac. Routine commands run on their own; a risky one waits for you.
Review and merge from anywhere
The pull request shows up on your phone and in the editor's Agents view. Read the diff, send a follow-up if it needs another pass, then approve and merge.
The pull request comes to you.
You don't have to be at the Mac that ran it. Celiador for iPhone lists what needs you, what is running and what is ready for review, with the agent's account of the change and the diff. The same runs sit in the editor's Agents view, and the issue on GitHub carries the record: picked up, the branch, the pull request.
- What the agent did, the diff, and what needs you, on your phone
- A follow-up starts another turn on the same branch
- Allow or deny a risky command without opening a laptop
- Approve and merge from the phone; the editor merges with your GitHub sign-in
Added an ApplePay.authorize call in checkout.ts and wired the token into capture(). Lint and tests pass locally.
Free while we're in early access.
The hosted agent runs on our key, with per-account limits we publish and enforce. Paid plans get a dated announcement before anyone is billed.
Early access
The whole product while we're in early access: the editor, the hosted Claude agent on our key, and the iPhone companion.
- Native macOS editor built on VS Code
- Hosted Claude agent — no key to bring, no card to add
- Daily limits published below and metered server-side
- iPhone companion to watch and steer runs
- Support answered by a person
Pro
When paid plans arrive, the price, the limits and the date appear here and in the changelog before a single invoice. Early-access accounts get the transition terms in writing, with an end date.
- Price and limits published before billing starts
- Self-service cancellation; statutory withdrawal rights honored
- No "unlimited" unless it is
Teams
Shared billing, seats and admin controls aren't built yet. If you'd run Celiador for a team, tell us what you'd need and we'll say plainly what exists.
- Nothing to buy today
- Your requirements shape what we build first
Full comparison on the pricing page.
Questions, answered
Celiador turns GitHub issues into pull requests. You label an 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 agent is Claude, run through our server on our key.
No. The Celiador editor, a fork of VS Code, has the loop built in and is the nicest place to watch runs. If you want to keep VS Code or Cursor, install the Celiador Remote extension: same loop, same account, same phone. Cloud runs for when your Mac is off are being built and will need the Celiador GitHub App on the repository.
Routine commands run on their own. A risky command or a question shows up under Needs you on the phone and in the Agents view, with Allow and Deny, and the run waits. Posting the same question on the issue, so you can answer by replying, is being built.
Yes, as early access. The editor is a signed, notarized macOS download that updates itself, with Linux x64 packages too; Celiador Remote is on Open VSX; the iPhone app is on TestFlight by invite until the public link is on the download page.
Runners: the Celiador editor on macOS (Apple Silicon) and Linux x64, or any VS Code or Cursor install with Celiador Remote. Review: iPhone, or the Agents view in the editor. A Windows build of the editor is on the roadmap; tell us on the download page if you need it.
Nothing during early access. The hosted agent runs on our key with per-account limits that are published on the pricing page and enforced by the server, and the app shows you the same numbers. Paid plans will be announced with a date before anyone is billed.
Your code stays yours. Runs happen on your machine, in a worktree of your own clone. It is never used to train models. No client holds a model key; the agent's requests go through our server, which holds the key, and nothing is indexed server-side.
Start with one issue.
Pick a runner, label an issue, and review the pull request from your phone. About ten minutes from install to the first merge.