Agent mode
Plans the change, edits several files, runs the tests and fixes what fails.
- Read orders.service.ts and its callers
- Updated 3 files
- Ran tests — 12 passed
The strongest Anthropic models for code — agent mode, code review and the CLI in VS Code and JetBrains. Officially through GitHub Copilot Business, with a contract and invoice from NQ DATA.
Plans the change, edits several files, runs the tests and fixes what fails.
Copilot reviews pull requests and proposes concrete fixes.
Copilot CLI explains, plans and runs tasks right in the repository.
Fable and Opus for hard problems, Sonnet for everyday work, Haiku for speed.
Admins choose models, set spending limits and exclude sensitive files.
| Model | Tier | Business | Enterprise |
|---|---|---|---|
| frontier | ✓ | ✓ | |
| deep reasoning | ✓ | ✓ | |
| deep reasoning | ✓ | ✓ | |
| everyday | ✓ | ✓ | |
| fast | ✓ | ✓ |
* Fable: enabled by the administrator, see “Data”. Model availability is set by GitHub and may change — we check it for your organisation before the contract.
1,900 AI credits per developer, pooled across the company
3,900 AI credits per developer, priority access to new models, requires GitHub Enterprise Cloud
Code completions and next-edit suggestions are unlimited. Chat and agents use credits by tokens; above the pool a credit costs $0.01 — we set limits so the invoice has no surprises.
Claude models in GitHub Copilot run with zero data retention. The exception is Fable 5 and 5.1: by default Anthropic retains prompts and outputs for safety classifiers; zero retention can be requested through GitHub until the end of 2026.
Yes. Administrators enable or disable each model and set policies for the organisation.
Additional usage is billed at $0.01 per credit if the administrator allows it; otherwise usage stops at the limit. Code completions are not affected.
For Copilot Business — no: it works with any GitHub organisation plan. Copilot Enterprise requires GitHub Enterprise Cloud.
Get an offer in one minute, or ask for a pilot: we connect 5–10 developers, configure models and limits, and review the results with you.
Choose the number of developers — the PDF offer is ready immediately.
Tell us about your team — we will suggest the plan, models and limits and launch a pilot for 5–10 developers.
Our team will contact you within one business day.