Forseti vs Cursor Bugbot
A fair, sourced comparison: where Forseti is stronger, where Bugbot is stronger, and what carries over if you switch.
Cursor Bugbot details accessed 15 Oct 2026
Who Bugbot suits
Cursor Bugbot, in fairness.
Cursor Bugbot suits teams that already work in Cursor and want review, fixes and post-deploy checks from one vendor. It reviews every push across five code hosts, including self-hosted GitLab and Bitbucket Data Center, and its Autofix agents can test a fix in a cloud VM and push it. Like Forseti, it reads its config from the base branch, so a PR cannot weaken its own review.[CB1][CB5]
Competitor details from their public pages, accessed 15 Oct 2026. Products change; tell us if something is out of date.
Feature table
| Feature | Forseti | Cursor Bugbot |
|---|---|---|
| Code hosts live | GitHub, live. GitLab, Bitbucket and Azure DevOps: adapters built, reviews switch on in a coming release. | GitHub, GitHub Enterprise Server, GitLab (including self-hosted), Bitbucket (including Data Center) and Azure DevOps.[CB1] |
| 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. | Usage-based since 11 May 2026: no seat fee, about $1.00–1.50 per run (previously $40 per seat). Needs at least one paid Cursor seat.[CB2][CB3] |
| Price shown before a review runs | Yes. Depth and price are posted on the PR before anything is spent. | Not stated publicly: there is no published rate card for effort levels.[CB4] |
| 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. | Not stated publicly.[CB3] |
| 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 Low, Default, High and Smart.[CB1] |
| 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. | Incremental review on each update; runs on every push by default.[CB1] |
| Verification step | A verification pass checks each finding before it is posted. Molecular depth (early access) adds sandboxed proof. | Autofix: cloud agents in VMs test and push fixes, up to 3 attempts.[CB5] |
| 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. | Nested .cursor/BUGBOT.md files; bugbot.yaml is read from the base branch.[CB1] |
| 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). | Learned rules and @cursor remember, with acceptance analytics per rule.[CB1] |
| 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. | Privacy Mode available. SOC 2, ISO 27001 and ISO 42001.[CB1] |
| Self-host or bring your own key | Not yet. Bring-your-own key and self-hosting are on the roadmap. | Not stated publicly.[CB1] |
| IDE and CLI | Not yet. A CLI and MCP pre-push verdict is coming. | Part of Cursor. An admin API with dry-run.[CB1] |
- Code hosts live
- ForsetiGitHub, live. GitLab, Bitbucket and Azure DevOps: adapters built, reviews switch on in a coming release.
- BugbotGitHub, GitHub Enterprise Server, GitLab (including self-hosted), Bitbucket (including Data Center) and Azure DevOps.[CB1]
- 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.
- BugbotUsage-based since 11 May 2026: no seat fee, about $1.00–1.50 per run (previously $40 per seat). Needs at least one paid Cursor seat.[CB2][CB3]
- Price shown before a review runs
- ForsetiYes. Depth and price are posted on the PR before anything is spent.
- BugbotNot stated publicly: there is no published rate card for effort levels.[CB4]
- 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.
- BugbotNot stated publicly.[CB3]
- 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. - BugbotEffort levels Low, Default, High and Smart.[CB1]
- 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.
- BugbotIncremental review on each update; runs on every push by default.[CB1]
- Verification step
- ForsetiA verification pass checks each finding before it is posted. Molecular depth (early access) adds sandboxed proof.
- BugbotAutofix: cloud agents in VMs test and push fixes, up to 3 attempts.[CB5]
- 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. - BugbotNested
.cursor/BUGBOT.mdfiles;bugbot.yamlis read from the base branch.[CB1] - 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).
- BugbotLearned rules and
@cursor remember, with acceptance analytics per rule.[CB1] - 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.
- BugbotPrivacy Mode available. SOC 2, ISO 27001 and ISO 42001.[CB1]
- Self-host or bring your own key
- ForsetiNot yet. Bring-your-own key and self-hosting are on the roadmap.
- BugbotNot stated publicly.[CB1]
- IDE and CLI
- ForsetiNot yet. A CLI and MCP pre-push verdict is coming.
- BugbotPart of Cursor. An admin API with dry-run.[CB1]
Bugbot 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 published price, quoted per PRForseti publishes a price per depth and posts it on the PR before the run. Cursor publishes no rate card for Bugbot’s effort levels.[CB4]
- Pushes do not multiply the billRe-reviews charge only for the new changes. Bugbot runs on every push by default, at about $1.00–1.50 a run.[CB1][CB2]
- Caps that degrade, never stopAt a cap the review drops to Glance and says so on the PR, so every PR still gets a review.
- No editor seat requiredForseti works with whatever editor your team uses. Bugbot needs at least one paid Cursor seat.[CB2]
- One record per PRThe Review Ledger keeps every finding’s ID and state across pushes, with one verdict per head SHA.
Where Bugbot is stronger
Credit where it is due.
- Five code hosts liveIncluding self-hosted GitLab and Bitbucket Data Center. Forseti is live on GitHub today; support for the other hosts is coming.[CB1]
- Autofix that pushesCloud agents test a fix in a VM and push it, up to 3 attempts. Forseti suggests fixes for a human to commit.[CB5]
- Security Review and RolloutsA separate security reviewer and post-deploy verification, both from 23 Sep 2026.[CB6]
- CertificationsSOC 2, ISO 27001 and ISO 42001. Forseti has none yet.[CB1]
- Acceptance analytics per ruleLearned rules come with acceptance numbers, and an admin API with dry-run.[CB1]
Switching
Switching from Bugbot.
Forseti imports your Bugbot 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
BUGBOT.md.cursor/BUGBOT.md.cursor/rulesWhat carries over
- Each
BUGBOT.mdbecomes team rules scoped to the folder it lives in, so nested files keep their reach .cursor/rulesfiles keep theirglobsas path-scoped rules- Your base-branch protection carries over: Forseti also reads rules from the base branch
What does not
- Effort settings (set depth with Forseti policies instead)
- Autofix and Rollouts settings
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
Cursor Bugbot sources
- CB1Cursor docs: Bugbot, accessed 15 Oct 2026
- CB2Cursor blog: May 2026 Bugbot changes, published 11 May 2026, accessed 15 Oct 2026
- CB3Cursor help: Bugbot usage-based billing, accessed 15 Oct 2026
- CB4Cursor forum: rate card for Bugbot not visible, published 12 May 2026, accessed 15 Oct 2026
- CB5Cursor blog: Bugbot Autofix, published 26 Feb 2026, accessed 15 Oct 2026
- CB6Cursor changelog: Rollouts and security reviewer, published 23 Sep 2026, accessed 15 Oct 2026