Member-only story
Parallel Coding Agents Are Easy. Coordinating Them Across Repos Isn’t.
How GitHub Copilot projects turn one goal into parallel sessions and a single review queue, with the architecture, guardrails, and costs behind it.
A change that touches an API, a web client, and a shared library means three agent sessions, three branches, and a plan that lives only in your head. The agents run in parallel; you still coordinate them one by one. GitHub Copilot projects move that coordination into the tool: you give Copilot one goal and the repositories it needs, Copilot runs parallel sessions toward that goal, and decisions, reviews, and pull requests come back to one queue. Below are the five layers behind it, where sessions run, the guardrails, the cost, and a pilot plan.
Copilot projects vs. agent sessions
A session is one task in one workspace. The GitHub Copilot app already runs many at once, each with its own git worktree and branch, and because the app is built on Copilot CLI with native GitHub integration, issues, pull requests, and CI checks work out of the box…
