codecloud
HomeDocsBlog
All posts

Ship features directly from Linear with our new native integration

January 26, 20264 min read

Codecloud now has a native Linear integration. Assign a ticket to @codecloud, and an agent implements it and opens a PR. No API calls, no n8n workflows, no middleman.

The previous approach—wiring up Linear webhooks through n8n or Zapier—worked, but required setup and didn't give you visibility into what the agent was doing. This integration fixes both problems.

How it works

Assign an issue to @codecloud in Linear. That's it. The agent receives the ticket via webhook and starts working.

The agent:

  1. Clones your repo and reads the ticket
  2. Posts its thinking to the issue comments as it works
  3. Implements the changes and runs any tests
  4. Opens a PR on GitHub when done

You can follow along in the Linear issue. The agent posts updates as it explores the codebase, makes decisions, and runs into problems. If something looks off, you can course-correct before it finishes.

Linear issue with codecloud agent posting progress updates

Talking to the agent

The agent reads issue comments. If it's going in the wrong direction, post a comment with more context. It'll pick it up and adjust.

Post "stop" to kill the session immediately. Useful if you realize the ticket wasn't ready or the agent is stuck in a loop.

Conversation between user and codecloud agent in Linear comments

Setup

In your codecloud dashboard, go to project settings → Integrations → Connect Linear. Authorize the OAuth app, select which Linear teams to enable, and you're done.

@codecloud shows up as a team member in Linear. Assign issues to it like you would any other teammate.

When to use this vs the API

The native integration is simpler for one-off tickets. Assign and forget. Good for clearing backlogs, quick bug fixes, or letting non-technical teammates trigger agent runs.

The API is better for automation at scale—CI/CD pipelines, custom workflows, or when you need programmatic control over model selection and parameters. The n8n approach from our earlier post still works if you want that level of control.

Ticket quality still matters

Same caveat as always: vague tickets produce vague results. "Fix the bug" won't get you far. Include what should change, where to look, and any constraints. The agent reads exactly what you write.


For setup details, see the documentation. Questions or issues? Use the support portal.