Relay

v0.2.0 ยท macOS and Linux

A project manager that works in your repos

Relay listens to Slack, Linear and ClickUp, answers from your code, and makes the changes you approve, on your machine and your own AI plan.

Relay's All activity list with requests from Slack and Linear

From message to merged, with you in charge

  1. 01

    Listens

    Slack, Linear, ClickUp and its own chat. Every message is saved before anything else happens.

  2. 02

    Understands

    One model call works out what's asked, which workspace and repo it's about, and how risky it is.

  3. 03

    Asks

    Questions get answered. Code changes and deletes wait for your approval, in the app or in Slack.

  4. 04

    Acts

    Approved changes land on their own branch, with a commit, and a pull request when you say so.

Answers

Answers from your code, in the thread

Ask in Slack, on a ticket, or in the app. Relay reads the repo read-only and replies where you asked, with the files it found.

A workspace chat answering where push notification permission is requested

Approvals

Nothing changes until you approve it

Relay drafts a plan and waits. Approve it and the change happens on its own branch. Approve again to push and open a pull request.

A plan to fix Android sign-in waiting for approval

Actions

Build and ship from a message

Saved actions build your apps and upload them to App Store Connect and Google Play, from a button, a chat message or a schedule.

The Actions tab with suggestions detected from a Flutter repo

Built to be trusted with your code

Runs on your plan

Uses Claude Code or Codex already signed in on your computer, and their Slack, Linear and ClickUp connections. No API keys needed.

Your checkout stays yours

A run won't start over uncommitted work. It uses its own branch, never commits keys, and switches back when done.

Nothing lost, nothing twice

Events are stored before cursors move. Replies go out through a durable outbox with retries.

Any model

Claude Code, Codex, Cursor and more on your plan, or Anthropic, OpenAI, Google, Mistral, Groq, xAI, OpenRouter and local models.

Only you can reach it

Listens on 127.0.0.1 with a token only your account can read. Replies are checked for keys before they're posted.

Guardrails built in

Replies start off. A daily token budget, a confidence threshold, and a log of every reply, approval and commit.

Start with what you already have

Install the app, point it at a repo, and ask it something. Slack, Linear and ClickUp can come later.