GitHub’s Copilot cloud agent is now generally available for Linear. As of July 23, you can assign a Linear issue to Copilot and walk away. It analyzes the issue, spins up an ephemeral GitHub Actions environment, writes the code, opens a draft PR, and streams progress back to your Linear timeline — no GitHub tab required. The GA release ships four new controls the preview lacked: model selection, custom agents, branch targeting, and live steering via comments. If you use Linear and GitHub together, this changes your workflow today.
How the agent loop works
The flow from issue to PR is fully autonomous. Assign an issue to Copilot in Linear, and the agent reads the issue, researches your codebase, creates an implementation plan, and executes it inside an isolated GitHub Actions environment. It runs your tests and linters, opens a draft pull request, streams progress updates back to your Linear activity timeline, and requests a review when it is done. You only re-enter the loop at the end, when a human decision is actually needed.
Two hard constraints are worth knowing upfront. Sessions time out at 59 minutes with no extension. The agent works within one repository and opens one PR per task — multi-repo changes are not possible. Both constraints push you toward smaller, well-scoped issues.
What the GA release actually adds
The preview has been live since October 2025. GA is not just a stability label — it ships meaningful controls:
- Model selection. Choose which model powers each task directly from your Linear workspace. The supported lineup includes Claude Sonnet, Opus, and Haiku variants, the GPT-5 family, and Gemini 3.6 Flash. Auto-select mode lets Copilot pick the best available model dynamically, which also carries a 10% AI credit discount.
- Custom agents. Point Copilot at a repository-specific agent that already knows your codebase conventions, test patterns, and deployment standards instead of starting cold.
- Branch control. Set the base branch and working branch per issue or workspace-wide. This matters for teams with protected main branches who route work through feature or staging branches.
- Live steering. Mention Copilot in a Linear comment while it is working to redirect it. If the initial plan is heading in the wrong direction, you can course-correct without killing the session.
What it handles well — and what it does not
The honest answer: write a good issue and the agent earns its keep. Write a vague one and you will spend more time fixing the PR than you would have spent writing the code.
The agent performs reliably on bug fixes with clear reproduction steps, test coverage expansion with explicit framework examples, refactoring within a defined scope, routine dependency upgrades, and documentation updates. These share one characteristic: the correct outcome is unambiguous before the session starts.
What breaks the agent is ambiguity. It cannot ask clarifying questions during execution. If the issue says “improve performance,” Copilot guesses what that means. Greenfield features with evolving requirements, cross-repo changes, and intermittent environment-specific bugs all land outside what a 59-minute autonomous session handles well.
Treat issue-writing as the skill. Specify file paths, not module names. Define acceptance criteria as explicit test cases. Link to relevant schemas and API contracts. Keep scope to two-to-four hours of equivalent human effort. That input quality directly determines whether the PR is usable or needs a rewrite.
Billing: AI credits plus Actions minutes
Copilot moved to usage-based billing in June 2026. Cloud agent sessions draw from two buckets: AI credits for token consumption and GitHub Actions minutes for the execution environment. Simple tasks run 5–15 premium requests; complex tasks can exceed 100. At $0.04 per premium request beyond your plan’s monthly allocation, a poorly specified complex task costs real money.
Monthly credit allocations by plan: Pro (1,500 credits), Pro+ (7,000), Business (1,900 per user), Enterprise (3,900 per user). Auto model selection offers a 10% discount on model multipliers if you do not need to pin a specific model.
Setup
Copilot cloud agent for Linear is available on all paid Copilot plans — Pro, Pro+, Business, and Enterprise. Install the GitHub Copilot for Linear app from the GitHub Marketplace. You need organization owner permissions on GitHub and workspace admin privileges in Linear. For Business and Enterprise plans, an administrator must enable the feature; it is off by default and can be scoped to specific repositories.
The integration docs on Linear’s site walk through the connection steps. Once enabled, the Copilot assignee option appears alongside your human teammates in the Linear issue interface.
The bigger shift
Copilot’s cloud agent is not the most capable autonomous coding agent available in July 2026 — Claude Code’s parallel subagents and one-million-token context window outperform it on complex tasks. What the Linear integration offers is something different: a genuinely asynchronous cross-platform loop that does not require you to babysit a terminal session or keep an IDE open. You write the issue, assign it, close the tab, and review a PR later. For teams that live in Linear, that workflow matches how work actually gets done.
The 59-minute constraint and single-repo limit are real. But for the tasks that fit inside those boundaries — which is a large fraction of day-to-day engineering work — the friction drop is significant. The question is not whether the agent is perfect. It is whether your issues are good enough to use it.
