GitHub opens parallel coding agents, drops per-seat Copilot billing for enterprises

Key Points

1. GitHub unveiled Copilot Projects, Fusion model routing and computer use at Universe 2026.
2. Enterprise customers get all Copilot features without a separate subscription, paying for AI credits consumed.
3. Agent-started pull requests hit 40% in September 2026, up from 17% in March.

The latest:

Software teams can now run coding agents across multiple repositories in parallel managed sessions, under a Copilot Projects public preview that GitHub announced on October 28, 2026 at GitHub Universe. The company bundled the release with a new multi-model orchestration layer, desktop computer control, and a billing change that lets GitHub Enterprise customers use every Copilot feature without a separate seat-based subscription.

Details:

• Copilot Projects: Available in the GitHub Copilot app from October 28, 2026, the preview lets a user describe a goal, attach context from repositories, design mockups, Slack threads and meeting notes, then pick models and tools. Copilot returns results such as pull requests for review, with one place to track progress and decisions.
• Copilot Fusion: A multi-model system that assesses what a task needs — reasoning, code generation, debugging, tool use — then routes execution across one or more models. Unite.AI and DEV Community reported it reached general availability after the HydraFusion preview; GitHub’s changelog says it supports local models.
• The benchmark: GitHub said the Fusion Max tier scored within two points of Opus 5.5 Max on the Artificial Analysis coding-agent index, at half the cost and 40% less time per task. A separate Copilot code review Max tier posted the highest F1 score on ReviewBench for high-severity issues, with Lite and Balanced tiers remaining.
• Computer use: Now in public preview on macOS and Windows, Copilot can read application content, click controls, enter text and move between apps — including ones with no API, command line or MCP integration. It requests approval before controlling any application, and organizations can disable it through managed settings.
• The billing shift: GitHub Enterprise customers can use the Copilot app, code review and automations without a separate Copilot subscription, paying instead for AI credits consumed. The option runs through sales now and self-service later. Copilot Business and Enterprise customers may stay or move. GitHub has not published credit pricing.
• Also shipped: Session continuity lets a cloud agent session carry across the GitHub app, VS Code, the command line and the web. Copilot Automations are repository workflows triggered manually, by event or on a schedule for tasks like issue triage, and now learn from outcomes to suggest improvements users approve. GitHub Agentic Workflows reached general availability.
• The Octoverse numbers: GitHub’s annual report, covering October 1, 2025 to September 30, 2026, found agents started more than 40% of pull requests in September 2026 against 17% in March, out of roughly 110 million requests that month. The platform logged 34.3 billion commits versus 12.9 billion in 2025, and 1.3 billion merged pull requests.
• Developer behavior: Paid users ran 47 different models in September against 29 in May; 61% of those choosing models used more than one and 63% used automatic selection at least once. Monthly active Copilot users touching agentic features reached 55% from 35% a year earlier, and cloud agent use grew 230%.
• The population: GitHub counted more than 225 million developers — 38.3 million in the United States, 32.0 million in India, 12.4 million in China and 10.2 million in Brazil. TypeScript ranked first among languages, followed by Python and JavaScript, with Rust the fastest growing in relative terms, nearly doubling.
• The infrastructure: GitHub said it added 4 million processor cores, 160 petabytes of storage and 50 terabits per second of network capacity this year. A move to multi-region deployment on Azure is expected to complete early next year.

Background:

The releases were announced at GitHub Universe 2026 and on the company blog under the byline of Jay Parekh, Microsoft corporate vice president for CoreAI. Parekh predicted that all software written to date will amount to less than 1% of what gets written in the next five years.

Between the lines:

Three announcements point the same way. Consumption billing replaces per-seat pricing precisely as agents, not humans, start the majority of pull requests — a seat count stops tracking usage once a single developer runs parallel sessions. GitHub’s own capacity additions suggest it is provisioning for that curve rather than for headcount growth.

What’s next

Credit pricing for the enterprise bundle remains unpublished, and self-service access has no announced date. The multi-region Azure deployment is due early next year, and Copilot Projects and computer use stay in public preview until general availability.