Today Aevral's pull-request reviews look for access control, business logic, injection, XSS, SSRF, path traversal, and LLM-integration security; whole-repo scans read access control and business logic. Claude Security is in public beta for Claude Enterprise. A finding is a lead, not a confirmation.
Aevral is an AI security reviewer for GitHub pull requests. It reviews each supported pull request for security flaws and suggests a fix; a human decides what merges. Built by Better ISMS, the company behind ISMS Copilot, in Paris, France.
Platform: GitHub only, through the Aevral GitHub App. GitLab, Bitbucket, and Azure DevOps are not supported.
Languages, PR review: PR review reads the diff of changed files in any language. Secret-bearing files (.env, private keys, credential files) and binaries are withheld, and very large pull requests keep the most security-relevant files.
Files are ordered security-relevant first, and the most security-relevant are also read in full at head. Generated files, lockfiles, and locale and fixture data are skipped by rule, and the review names them. The review is not tuned per language.
Languages, whole-repo scan: The scan reads source files in TypeScript, JavaScript, Python, Go, Ruby, PHP, Java, Kotlin, Rust, C#, Swift, Scala, Vue, Svelte, SQL, and GraphQL. Other file types, tests, and vendored or build folders are skipped.
Scope limits: Not a secret scanner, not dependency or SCA scanning, not memory corruption, not a general SAST.
Teams that do not work through pull requests on GitHub.
Enterprise AppSec buyers who need demos, SSO, or a procurement process first: Aevral is self-serve, and the Enterprise console features (SSO / SAML, RBAC, audit log) are not shipped.
Buyers who need compliance evidence on pull requests (ISO 27001, SOC 2): that is heyGRC, from the same company, https://heygrc.com.
Open-source maintainers are welcome: public repositories are free, 500 reviews per organization per month.
Self-serve is live for PR security review and for whole-repo scans.
PR review starts when the GitHub App is installed and looks for access control, business logic, injection, XSS, SSRF, path traversal, and other flaws on the diff; the scan reads authorization, IDOR, and business-logic access control.
Install the GitHub App, then sign in to the console with GitHub (it connects the install automatically), and run a scan. Hosted Claude Security scans are in public beta for Claude Enterprise.
The Claude Security plugin is in beta for Claude Code users.
No. Aevral is not open source. The engine, the planted-vulnerability corpus, and the judge stay private. What is public: the transparency report, the methodology page, the agent-setup prompt, and the plugin repo. How the scan works, in prose: https://docs.aevral.com/docs/how-aevral-hunts.
Aevral is an AI security reviewer for GitHub pull requests: scan the repository, review pull requests, suggested fixes. Open-source models, processed in the US today, EU-only option announced. Named competitors: Claude Security and Codex Security. See https://aevral.com/compare/claude-security and https://aevral.com/compare/codex-security.
Two surfaces.
PR review looks for access control, business logic, SQL and command injection, XSS, SSRF, path traversal, unsafe deserialization, token and session flaws, and LLM-integration risks on the diff: findings are leads, advisory, at most five per review, and on a large pull request the most security-relevant files are reviewed first (the review says which files it covered).
Whole-repo scans read authorization, IDOR, and business-logic access control only. Not memory corruption, not secret or dependency scanning, not a general SAST. Languages are listed under What does Aevral support?
No. Aevral is an AI security review with mechanical gates, not a rules engine. Each review reads the diff of every changed file plus the full head content of security-critical changes; it never reads git history and never executes code.
A published finding must name the attacker, the input, the path, and the missing control, and its cited code must match, verbatim, lines the pull request actually added, a check the engine runs mechanically against the diff.
Rule-based SAST stays yours: deterministic whole-tree scanning, secrets, dependencies, memory classes. Keep both. The comparison: https://aevral.com/is-aevral-sast. The mechanics: https://aevral.com/how-it-reviews.
No. Detection is probabilistic: runs vary, misses happen, and both facts are public.
The gates around the model (the publication contract, the verbatim grounding check, the proof-required verdict, the cap of five findings, silence on clean pull requests) are deterministic and live in code.
Planted-vulnerability runs with the misses included are published on the receipts page: https://aevral.com/receipts.
The whole-repo scan class list above is what the scan covers today; broader scan classes are not sold. If that list shrinks further, customers are told and must opt in again.
Two products, both priced per organization, excluding VAT.
Whole-repo scans, USD: Public repositories USD 0 (1 authorized public-repo scan a month), Team USD 99 (4 scans, then $29), Business USD 399 (16, then $19), Scale USD 1,699 (100, then $17), Enterprise by quote from USD 3,300 per month equivalent.
PR security review, USD: the free tier is live (public repos free, 500/org/month, 25 private reviews a month from install; a 14-day, 500-review trial starts when you sign in with GitHub and connect the install); paid plans from $49 a month are live in the console.
The billable-scan definition, overage opt-in and the spend cap are shown before purchase. See https://aevral.com/pricing.
The Public repositories plan includes 1 scan of an authorized public repo per calendar month. Aevral itself is not open source. You must have authorization to scan, and findings are not auto-published.
Sometimes: a finding can carry a suggested fix as a diff you apply, and comes with a fix prompt for Claude Code, Cursor, or Codex. Nothing is applied without you. Aevral does not commit, push, or auto-merge patches.
PR review posts an advisory Check and comments when there is a finding.
A finding is a lead with evidence. Review it before you change production code. There is no confirmation pass yet; that is on the product roadmap. Do not treat a report as a clean bill of health.
Yes. Scans can be wrong. Review every finding and every fix prompt before you change production code, especially on critical systems.
Not necessarily. Scans are stochastic. Re-runs can differ. That is a property of this kind of analysis, not a clean-bill ritual.
No. No finding does not mean secure. An incomplete scan is not a clean bill of health.
No. Aevral does not replace SAST, SCA, secrets detection, or a pentest. It is a whole-repo security scan aimed at the classes above.
Whole-repo means the eligible files in the snapshot we scan, starting on the default branch. It is not a promise that every file, generated artifact, or submodule is in scope.
The family trust center is https://trust.ismscopilot.com. Better ISMS is getting ISO 27001 certified. The intended scope includes how we build and run Aevral, ISMS Copilot, and heyGRC. We are not certified today. Aevral is not product-certified.
Read the intended scope at https://trust.ismscopilot.com/en/iso-27001.
PR security review is Aevral's lead product: a reviewer on your pull requests that posts a Check plus inline comments grounded on the added lines, looking for access control, business logic, injection, XSS, SSRF, path traversal, and LLM-integration security.
PR review is live on install: installing the App starts reviews on, reviews can start on the next pull request even before anyone signs in, and the owner can turn them off. See https://aevral.com/products/pr-review.
Public repositories are free, 500 reviews per organization per month. Reviews start at install: before anyone signs in, 25 private reviews a month are free.
Signing in with GitHub connects the install and starts a 14-day trial: private reviews are free up to 500, and the trial ends at 14 days or 500 reviews, whichever comes first.
After the trial, 25 private reviews a month are free (trial-covered reviews do not consume the subsequent Free allowance); past that, a pull request gets a neutral Check and no review until the next month. No card.
Paid plans are purchasable in the console.
Paid plans: Starter $49 / 100 PR reviews a month, Pro $199 / 500, Business $999 / 3,000, Scale $1,999 / 6,500, USD per organization, monthly.
Business adds a monthly export of the security reviews Aevral ran on your pull requests, with coverage and results, that may serve as supporting evidence for automated security testing in audits (for example ISO 27001 A.8.29, SOC 2 CC8.1), and EU-only processing when it is available; Scale carries everything in Business.
Enterprise starts from $4,000 a month and is not self-serve. A standard DPA is available on every paid plan. Paid PR plans are live in the console.
On paid plans a PR review is counted per pull request per billing period; a re-push does not count again.
The first time an organization goes past its plan, it gets 25% more PR reviews free, once; after that, $0.99 per extra PR review on every paid tier if the owner chose extras, otherwise reviews pause until the next period.
Scale at its limit: extras or Enterprise. A review is never one of the included whole-repo scans of a scan plan, and no scan plan is required.
No live MCP endpoint today. A coding-agent plugin is public: https://github.com/better-isms/aevral-plugin. PR security review, above, is live on install. The owner can turn reviews off. The scan fix path today is evidence plus a prompt you copy in the console.
The GitHub Check on the scanned commit is a count plus a link to that report, not a PR comment.
No. Aevral is not affiliated with, endorsed by, or sponsored by Anthropic. Claude and Claude Security are used nominatively to identify the product we compete with.
Install the GitHub App and PR review starts on. Sign in with GitHub to connect it, then press Scan for the repository you already have.
For professional use. By installing, you confirm you can act for the account or organization that owns it, and you accept the Terms and DPA on its behalf.