Forseti vs GitHub Copilot
A fair, sourced comparison: where Forseti is stronger, where Copilot is stronger, and what carries over if you switch.
GitHub Copilot details accessed 15 Oct 2026
Who Copilot suits
GitHub Copilot, in fairness.
GitHub Copilot code review suits teams that already pay for Copilot and want review inside GitHub with no extra vendor. It runs agentically on GitHub’s own runners, resolves its comments once they are addressed, hands fixes to the Copilot cloud agent, and is billed from the same AI credits as the rest of Copilot. GitHub reports 60 million Copilot code reviews.[GH1][GH2][GH7]
Competitor details from their public pages, accessed 15 Oct 2026. Products change; tell us if something is out of date.
Feature table
| Feature | Forseti | GitHub Copilot code review |
|---|---|---|
| Code hosts live | GitHub, live. GitLab, Bitbucket and Azure DevOps: adapters built, reviews switch on in a coming release. | GitHub. Azure DevOps in preview. No GitLab or Bitbucket.[GH1] |
| Pricing model and list price | Per PR, priced by depth. 1 credit = £0.01. A medium PR (about 1,000 lines): Glance £0.03, Standard £0.79, Deep £4.64. Team plan £8 per active PR author per month with 800 credits included. Recommended pricing, final after our golden-set evaluation. | AI Credits since 1 Jun 2026 ($0.01 each, charged at token rates). Plans include credits equal to their price: Business $19, Enterprise $39. Reviews on private repos also use GitHub Actions minutes.[GH4][GH3] |
| Price shown before a review runs | Yes. Depth and price are posted on the PR before anything is spent. | Not stated publicly. Cost follows the tokens a review uses, plus Actions minutes.[GH4][GH3] |
| When a spend limit is reached | Caps per PR, per developer and per month. At a cap the review drops to Glance and says so on the PR; it never stops silently. No auto-refill. | If the member’s credit quota runs out, the review fails.[GH1][GH4] |
| Depth control | Glance, Standard or Deep (Molecular in early access), set by policy per repo, branch and path, or with an @forseti command on the PR. | Effort levels Lite and Balanced (Max announced), set per request, requester, repo, org or enterprise. The default changed to Balanced on 28 Sep 2026. No path-based trigger.[GH5][GH6] |
| Finding lifecycle across pushes | Review Ledger: every finding keeps its ID and state across pushes, with one verdict per head SHA. Re-reviews charge only for the new changes. | Resolves its comments once they have been addressed.[GH1] |
| Verification step | A verification pass checks each finding before it is posted. Molecular depth (early access) adds sandboxed proof. | Agentic tool calling on Actions runners, with shell tools to validate.[GH2] |
| Config file, and where rules are read from | .forseti.yaml and .forseti/rules.md, read from the base branch, so a PR cannot change its own review. | copilot-instructions.md, *.instructions.md with applyTo globs, AGENTS.md and skills, read from the PR’s head branch.[GH1] |
| Learning from feedback | In testing: built and being verified. Team memory: learns what your team dismisses, and shows its precision (findings accepted against findings dismissed). | Copilot Memory (preview).[GH1] |
| Data retention and training | No repository or diff storage: patches are fetched live and held in memory for at most 2 minutes. Findings keep a short snippet unless you switch it off. We do not train on your code. | Not used for training on Business plans.[GH1] |
| Self-host or bring your own key | Not yet. Bring-your-own key and self-hosting are on the roadmap. | Not stated publicly.[GH1] |
| IDE and CLI | Not yet. A CLI and MCP pre-push verdict is coming. | Reviews can be requested through the API with an effort parameter (from 2 Oct 2026). Fixes hand off to the Copilot cloud agent.[GH6][GH1] |
- Code hosts live
- ForsetiGitHub, live. GitLab, Bitbucket and Azure DevOps: adapters built, reviews switch on in a coming release.
- CopilotGitHub. Azure DevOps in preview. No GitLab or Bitbucket.[GH1]
- Pricing model and list price
- ForsetiPer PR, priced by depth. 1 credit = £0.01. A medium PR (about 1,000 lines): Glance £0.03, Standard £0.79, Deep £4.64. Team plan £8 per active PR author per month with 800 credits included. Recommended pricing, final after our golden-set evaluation.
- CopilotAI Credits since 1 Jun 2026 ($0.01 each, charged at token rates). Plans include credits equal to their price: Business $19, Enterprise $39. Reviews on private repos also use GitHub Actions minutes.[GH4][GH3]
- Price shown before a review runs
- ForsetiYes. Depth and price are posted on the PR before anything is spent.
- CopilotNot stated publicly. Cost follows the tokens a review uses, plus Actions minutes.[GH4][GH3]
- When a spend limit is reached
- ForsetiCaps per PR, per developer and per month. At a cap the review drops to Glance and says so on the PR; it never stops silently. No auto-refill.
- CopilotIf the member’s credit quota runs out, the review fails.[GH1][GH4]
- Depth control
- ForsetiGlance, Standard or Deep (Molecular in early access), set by policy per repo, branch and path, or with an
@forseticommand on the PR. - CopilotEffort levels Lite and Balanced (Max announced), set per request, requester, repo, org or enterprise. The default changed to Balanced on 28 Sep 2026. No path-based trigger.[GH5][GH6]
- Finding lifecycle across pushes
- ForsetiReview Ledger: every finding keeps its ID and state across pushes, with one verdict per head SHA. Re-reviews charge only for the new changes.
- CopilotResolves its comments once they have been addressed.[GH1]
- Verification step
- ForsetiA verification pass checks each finding before it is posted. Molecular depth (early access) adds sandboxed proof.
- CopilotAgentic tool calling on Actions runners, with shell tools to validate.[GH2]
- Config file, and where rules are read from
- Forseti
.forseti.yamland.forseti/rules.md, read from the base branch, so a PR cannot change its own review. - Copilot
copilot-instructions.md,*.instructions.mdwithapplyToglobs, AGENTS.md and skills, read from the PR’s head branch.[GH1] - Learning from feedback
- ForsetiIn testing: built and being verified. Team memory: learns what your team dismisses, and shows its precision (findings accepted against findings dismissed).
- CopilotCopilot Memory (preview).[GH1]
- Data retention and training
- ForsetiNo repository or diff storage: patches are fetched live and held in memory for at most 2 minutes. Findings keep a short snippet unless you switch it off. We do not train on your code.
- CopilotNot used for training on Business plans.[GH1]
- Self-host or bring your own key
- ForsetiNot yet. Bring-your-own key and self-hosting are on the roadmap.
- CopilotNot stated publicly.[GH1]
Copilot prices are in US dollars as published. Forseti prices are our recommended pricing, final after our golden-set evaluation. “Not stated publicly” means we found no public statement on 15 Oct 2026.
Strengths
Where Forseti is stronger
Cost you can see, depth you choose.
- A PR cannot rewrite its own reviewForseti reads its rules from the base branch. Copilot reads its instructions from the PR’s head branch, so the PR under review can change them.[GH1]
- Out of credit still means reviewedAt a cap, Forseti drops to Glance and says so on the PR. When a member’s Copilot quota runs out, the review fails.[GH1]
- Depth by path, branch and repoDeep on
billing/**, Glance on docs. Copilot sets effort by request, user, repo, org or enterprise, with no path-based trigger.[GH5] - The price is on the PR before it runsA quoted price per depth, before anything is spent. Copilot bills tokens and, on private repos, Actions minutes.[GH4][GH3]
- Your CI minutes stay yoursForseti reviews on its own infrastructure; nothing runs in your Actions.
- One record per PRThe Review Ledger keeps every finding’s ID and state across pushes, with one verdict per head SHA.
Where Copilot is stronger
Credit where it is due.
- Already inside GitHubNo app to install and no new vendor for teams on Copilot.[GH1]
- Credits come with the planCopilot Business and Enterprise include AI credits equal to the plan price.[GH4]
- Fix hand-offA finding can go straight to the Copilot cloud agent, which opens a fix PR.[GH1]
- ScaleGitHub reports 60 million Copilot code reviews by March 2026.[GH7]
- Auto-approval by path (preview)Copilot approvals can approve a PR when every changed file matches up to 15 path globs.[GH1]
- Azure DevOps in previewBeyond GitHub, Azure DevOps support is in preview.[GH1]
Switching
Switching from Copilot.
Forseti imports your Copilot config, shows you what maps and what does not, and applies it when you say so. Paste the files, or let Forseti read them from your default branch.
Reads
.github/copilot-instructions.md.github/instructions/*.instructions.mdWhat carries over
.github/copilot-instructions.mdsections become team rules*.instructions.mdfiles keep theirapplyToglobs as path-scoped rules- From then on the rules are read from your base branch, not the PR’s head
What does not
- Effort settings (set depth with Forseti policies instead)
- Copilot Memory
The preview lists every unmapped key, so nothing is dropped silently. Import is audited, and you can export everything back as .forseti.yaml and .forseti/rules.md.
See the price on your next PR.
Install the GitHub App, pick a repo, and Forseti reviews a real PR with the price on it before it runs.
Sources
GitHub Copilot sources
- GH1GitHub Docs: Copilot code review, accessed 15 Oct 2026
- GH2GitHub changelog: agentic architecture, published 5 Mar 2026, accessed 15 Oct 2026
- GH3GitHub changelog: Actions minutes from 1 Jun 2026, published 27 Apr 2026, accessed 15 Oct 2026
- GH4GitHub changelog: Copilot billing and plans, published 1 Jun 2026, accessed 15 Oct 2026
- GH5GitHub changelog: effort levels GA, published 7 Aug 2026, accessed 15 Oct 2026
- GH6GitHub changelog: API support and new default effort, published 2 Oct 2026, accessed 15 Oct 2026
- GH7GitHub blog: 60 million Copilot code reviews, published 5 Mar 2026, accessed 15 Oct 2026