Skip to main content
AI-Implement authenticates to Linear as an app, not as a person. You create an app in your Linear workspace, enable client-credentials tokens on it, and give the orchestrator its client ID and secret. Because the orchestrator acts as the app, every status change, label, and comment it makes is attributed to the app rather than to whoever set up the credentials. Automated activity stays clearly distinguishable from your team’s, and the name you give the app is what appears on that activity.
This page applies to Linear. If your project mappings use Jira, configure the Jira credentials on the orchestrator instead — see Environment variables.

​
Before you start

You need the Admin role on the Linear workspace to create an application. Team Owner is not sufficient — that role is scoped to a team and cannot create workspace applications.If you don’t have Admin, ask someone who does to create the app and pass you the client ID and secret. Nothing else in this guide needs elevated access.
An app’s credentials work only on the workspace it was created in, so one orchestrator instance serves one Linear workspace. To connect a second workspace, create a second app there and run a second orchestrator instance against it.

​
Create the Linear app

1

Open the application settings

In Linear, go to Settings → API → Applications, then choose New application.
2

Fill in the application details

Linear requires at least one redirect URI on every application, even though the client-credentials flow never redirects a browser anywhere. A placeholder such as https://localhost satisfies the form and is never called.
The GitHub username field only affects how Linear’s own GitHub integration attributes commits and pull-request comments. AI-Implement links pull requests to issues itself, so you can leave it blank.
3

Enable client credentials

Turn on Client credentials. Leave Public and Webhooks off — all three are off by default.
This toggle is what lets the orchestrator mint a token. If you miss it, the app is created successfully but the orchestrator fails to authenticate at startup, which makes the cause easy to overlook.
4

Copy the client ID and secret

Both values stay available on the application’s settings page, so you can come back for them whenever you need to configure another environment.

​
Add the credentials to your orchestrator

Set these three variables on the orchestrator. For a Fly.io deployment use fly secrets set for the first two; for local development put all three in your .env file.
Set LINEAR_WORKSPACE_URL even though the orchestrator starts without it. It’s the base for every issue link in notifications and the admin UI, and it falls back to https://linear.app — so leaving it unset produces links that look valid but don’t resolve to your workspace.

​
Prepare the Ready for Review label

AI-Implement creates the per-team labels it needs automatically. One label is different: Ready for Review is workspace-level, and creating a workspace-level label requires admin permission that an application identity cannot hold. If your workspace already has a label with that name, AI-Implement reuses it and there’s nothing to do. If it doesn’t, create it by hand once before your first run.

​
What’s next

Deploy the orchestrator

Run the orchestrator on Fly.io or locally with the credentials you just created.

Connect a target repo

Sync workflow templates into a repo and map it to a Linear team.