Getting started

The loop

Label a GitHub issue celiador on a repository you have linked from the phone. A runner on your Mac picks it up, creates a git worktree on a fresh branch so your checkout is untouched, and runs a Claude agent there. The issue gets a comment saying it was picked up and on which branch. When the agent is done, the runner pushes the branch and opens a pull request that closes the issue; the issue gets the link. You review the diff on the phone or in the editor's Agents view, send a follow-up if it needs another turn, and approve and merge.

Runners: the Celiador editor (below), or VS Code and Cursor with the Celiador Remote extension. 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 waits for you under Needs you, with Allow and Deny. The start guide walks through the GitHub App, the phone and the first issue.

Install

The Celiador editor runs on Apple Silicon Macs with macOS 13 or later, and on Linux x64. Download the DMG, open it, drag Celiador to Applications, and open it from there. The app is signed with a Developer ID and notarized by Apple: you'll see macOS's standard “downloaded from the internet” confirmation once, and nothing else. Every build we ship, with checksums, is on the releases page.

Linux packages (.deb, .rpm, .tar.gz) are on the download page; Windows is on the roadmap and the same page has a notify-me form for it. Keeping VS Code or Cursor? Install Celiador Remote instead; the switching page has the steps.

Sign in

The first launch opens a welcome view with a Sign in button (it's also the account item in the status bar, and the command Celiador: Sign In). Pick email and password, a magic link, or GitHub. A magic link must be opened on the same Mac within ten minutes; it hands you straight back into the editor. Your account is what the hosted agent and the iPhone companion share.

The agent in the editor

Claude is built into the editor and runs through our server; there is no key to paste and no extension to install. Labelled issues run through it without you doing anything (see the loop above). The first chat downloads the agent runtime (about a minute), then three surfaces are yours for work you start yourself:

Chat

⌃⌘I opens the chat view (or the chat icon in the title bar). Ask about the codebase, or ask for a change: the agent plans, edits across files, runs your tools, and every write is a diff you accept or reject. Attach files with #file: or Add Context; the active editor is forwarded automatically.

Inline edit

⌘I in a file edits the selection or the file in place and shows the result as inline hunks: ⌘Y keeps, ⌘N undoes. ⌘I in the terminal asks for a command and offers to run or insert it; it never runs on its own.

Agents window

Open Agents Window from the Command Palette is the cockpit: the queue of labelled issues and several runs side by side, each with its transcript, tool calls and diff. These are the same sessions the iPhone app watches, and the pull requests can be reviewed and merged here as well as on the phone.

iPhone companion

The companion app is the review surface for the loop: what needs you, what is running, and the pull requests ready for review, from every runner on your account. Read the agent's account of the change and the diff, send follow-ups, allow or deny a tool call, answer the agent's questions, cancel, browse and edit files of the session, discard files, and approve and merge the pull request, without touching the laptop. “From your editor” lists the repositories your Mac has open, so a run can also start from the phone on the local working copy; no cloning, no tokens. Distribution is through TestFlight; the link appears on the download page when it's live.

Workspace Trust

A repository can carry settings, tasks and agent hooks that execute code. Celiador keeps VS Code's Workspace Trust on: a folder you haven't trusted opens in Restricted Mode, where its settings are filtered and its tasks don't run, and the agent will not start in it until you trust it. Trust is per folder and you can review or revoke it at any time (Workspaces: Manage Workspace Trust).

Even in a trusted folder, the agent cannot change what the editor or the agent itself executes without your explicit approval: edits to .vscode/*.json, agent settings under .claude/, git hooks and similar files always prompt, in every permission mode, including after you approved a plan.

Limits

The hosted agent runs on our key with per-account limits that the server enforces and the app shows you in its account panel. The current numbers, live from the same endpoint:

Live from celiador.ai/api/limits. Why we publish them: the conduct pledge.

Updates

The app checks celiador.ai for a newer build on its own, downloads it in the background, and offers a restart; Celiador: Check for Updates does it on demand. Each build is notarized before it's published, and the changelog says what changed.

Your data

Your code stays on your machine; the agent reads what it needs for a request and nothing is indexed server-side. Your code is never used to train models. The server stores your account, chat history and push registrations so the phone and the editor see the same sessions; deleting your account removes all of it. Details in the privacy policy.

Support

A person answers. Write to hello@celiador.ai or use the contact page; bugs and requests are welcome on GitHub.

For crawlers: the site publishes a /sitemap.xml and a /robots.txt.