# Aevral

Aevral is an AI security reviewer for GitHub pull requests. Security review, handled. Your agents open pull requests faster than anyone can read them. Without a reviewer, the security flaw still ships in the product you own. PR review looks for access control, business logic, injection, XSS, SSRF, path traversal, and LLM-integration flaws; the whole-repo scan reads authorization, IDOR, and business logic. A suggested fix you apply in Claude Code, Cursor, or Codex. Open-source models, processed in the US today, EU-only option announced. An alternative to Claude Security and Codex Security. The whole-repo scan is live, for code already in the repository. Priced by organization, not by seat. No floor to start. Built by Better ISMS, the company behind ISMS Copilot, in Paris, France.

## Products

Two products, two surfaces, two prices. Review is first. The whole-repo scan is live and bounded to access control.

1. PR security review: a Check on the head commit plus inline comments grounded on added lines, looking for access control, business logic, SQL and command injection, XSS, SSRF, path traversal, unsafe deserialization, token and session flaws, and LLM-integration risks; advisory, never blocks a merge; up to five findings per review. Needs Pull-requests read and write. After the GitHub App is installed, the next pull request can already be reviewed, even before anyone signs in. Setup Complete can still turn them off. An opt-out stays off while the App stays installed; reinstalling the App turns reviews back on. Paid PR plans live in the console.
2. Whole-repo scan (live, at rest): a reading of the default-branch snapshot of a repository. Trigger: the console Scan button or a paid recurring schedule. Output: a GitHub Check on the scanned commit, a console report with evidence, and a copy fix prompt. Today it scans authorization, IDOR, and business-logic access control only; the wider class list is PR review. Not memory corruption, not secret or dependency scanning, not a general SAST. GitHub App permissions for this product: Contents read, Metadata, Checks write.

## Supported / Not for

Supported:

- 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.

Not for:

- 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.

## Status

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. Aevral is self-serve: install the GitHub App (https://github.com/apps/aevral), then sign in to the console with GitHub; it connects the install automatically. Claude Security is in public beta for Claude Enterprise. Where the site still shows a waitlist CTA (https://tally.so/r/dWe0kK), joining does not reserve a price, access, or a contract. There is no marketing mailto. Contact form: https://tally.so/r/PdxdbV.

Present-tense verbs: review, scan, suggest a fix. Aevral does not validate findings. A finding can carry a suggested fix as a unified diff the buyer applies; Aevral never applies or merges it. Copy fix prompt into Claude Code, Cursor, or Codex (equal weight). A finding is a lead with evidence, not a confirmation. No patches are auto-applied. No live MCP endpoint today; do not claim one. After the GitHub App is installed, the next pull request can already be reviewed. Paid PR plans live in the console.

## Prices

Two products, both priced per organization, never per seat; neither requires the other. PR security review, live on install, USD per organization per month excluding VAT, no scan plan required (PR reviews start at install, before anyone signs in; the App install must have accepted Pull-requests write): public repositories free, 500 reviews per organization per month (they do not consume the private-review allowance; extra public reviews resume the next calendar month). Connecting the install in the console (GitHub sign-in) starts a 14-day trial with a 500-review cap; the trial ends at 14 days or 500 private reviews, whichever comes first. After the trial, 25 private reviews per calendar month (UTC) free (trial-covered reviews do not consume the subsequent Free allowance); at the cap a pull request gets a neutral Check and no review until the next month. No card. Paid plans live in the console: Starter $49 / 100 PR reviews a month, Pro $199 / 500, Business $999 / 3,000 (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 (not live yet)), Scale $1,999 / 6,500 (everything in Business), then $0.99 per extra review on every paid tier if the owner turned extras on, at checkout or later in the console, otherwise reviews pause until the next period with a Check saying why. The first time an organization goes past its plan it gets an extra quarter of the plan free, once, before any extra is billed. On paid plans a PR review is counted per pull request per billing period; re-pushes to that pull request do not count again. Scale at its limit: extras or Enterprise (from $4,000 a month, not self-serve, custom DPA; SSO when available). A standard DPA is available on every paid plan. The free tier has no extra reviews. On Free and the trial, a review is one pull request head (repository, PR number, head commit); the same head is never counted twice; reviews skipped or failed before posting are not counted. Fair use: hourly and daily review limits apply per organization, higher on paid plans. After the first review of a pull request, a re-push reviews only what changed since the last reviewed commit; a force-push, a rebase or a reopen gets a full review again. On a large pull request the most security-relevant files are reviewed first and the review says which files it covered; a review always appears. A review is never one of the included whole-repo scans of a scan plan.

Whole-repo scan, live, USD per organization per month, HT B2B excluding VAT: Public repositories USD 0 (1 authorized public-repo scan per calendar month per organization; a second public scan that month is refused), Team USD 99 (4 default-branch scans, then $29 per extra scan, opt-in), Business USD 399 (16 then $19), Scale USD 1,699 (100 then $17). Team, Business and Scale are self-serve in the console. Enterprise by quote: from USD 3,300 per month equivalent (200 scans and up, $16.50 per scan minimum), annual commitment, invoice billing; not self-serve, Enterprise console features (SSO/SAML, RBAC, audit log, multiple GitHub orgs, invoice/PO tooling) not shipped. A scan counts when newly admitted (re-scanning the same commit in a period returns the existing scan); a scan that ran counts whether or not it found anything; only a never-run scan is refunded. Included scans reset with the billing period (calendar month for Public repositories), no rollover. Extra scans are opt-in under an owner-set monthly cap on extra-scan spend.

Price shapes compared with Claude Security (org units vs Enterprise seats plus tokens; different units, do your own math): https://aevral.com/pricing/compare.

## For agents

Customer docs center (how-to corpus): https://docs.aevral.com/docs. Its machine index: https://docs.aevral.com/llms.txt. Agent hub on the docs host: https://docs.aevral.com/docs/for-ai-agents. Agent one-pager (GTM page): https://aevral.com/for-agents. Per-agent pages: https://aevral.com/for-agents/codex, https://aevral.com/for-agents/cursor. Agents whose work arrives as a pull request, each reviewed like any other pull request with a copy-paste fix prompt a human hands back (no integration): https://aevral.com/for-agents/claude-code, https://aevral.com/for-agents/github-copilot, https://aevral.com/for-agents/devin, https://aevral.com/for-agents/jules, https://aevral.com/for-agents/codex-cloud. Official agent-setup prompt: https://docs.aevral.com/agent-setup/prompt.md. Coding-agent plugin: https://github.com/better-isms/aevral-plugin. How Aevral hunts: the scan method behind the security review and one abridged prompt excerpt: https://docs.aevral.com/docs/how-aevral-hunts. aevr_ API keys exist today (console Developer API page, shown once): they start and read scans and read plan and usage for their own organization; billing is console-session-only. Usage API (GET https://aevral-worker-prod.fly.dev/v1/usage, read-only; the billing link it returns is for an owner or admin to open): https://docs.aevral.com/docs/usage-api. Live plan ladders as JSON, no key: PR review https://aevral-worker-prod.fly.dev/v1/pr/billing/plans, scans https://aevral-worker-prod.fly.dev/v1/billing/plans; agents quote prices only from these, never from memory. Scan API (POST https://aevral-worker-prod.fly.dev/v1/scans, GET /v1/scans/{id}; bearer aevr_ key read from the AEVRAL_API_KEY environment variable, never pasted in chat; poll status_url every 30s, scans take 10 to 40 minutes): https://docs.aevral.com/docs/scan-api. No live MCP endpoint today; do not claim one. This file is the durable machine-readable source of truth for products, prices, routes, and status. Full route index: https://aevral.com/llms-full.txt. AI crawler directives: https://aevral.com/ai.txt. Agent manifest: https://aevral.com/.well-known/agent.json.

## Compare

Compare hub: https://aevral.com/compare. Claude Security and Codex Security are named competitors, each with an even, dated nine-row table. Same jobs: scan the repository, review pull requests, and suggest fixes. Aevral is not a lab: it runs on open-source models, processed in the US today, with an EU-only option announced. Of the sixteen alongside tools, ten publish security scope that touches the job and six state a different job. These are product comparisons, not catch-rate claims. The complementary pages state each tool's own published job and never claim what that tool would or would not flag; Aevral adds its PR review alongside it, which looks for access control, business logic, injection, XSS, SSRF, path traversal, unsafe deserialization, token and session flaws, and LLM-integration flaws on the diff, while its whole-repo scan reads authorization, IDOR, and business-logic access control only. Aikido's platform (SAST, dependencies, secrets, IaC, pentest) is complementary. Aikido also publishes Code Security Audit and Deep Review for authorization, IDOR, and business-logic reading of source; Aevral's whole-repo scan is built for that reading; its PR review, live on install, looks for that and a wider class list on the diff. See https://aevral.com/compare/aikido. CodeRabbit also publishes CodeRabbit Security, agentic code security monitoring: scheduled repository scans and security scans of each pull request, per CodeRabbit's site. CodeRabbit Security and Aevral's pull-request review, live on install, both read pull requests for security. See https://aevral.com/compare/coderabbit. Greptile publishes security, including a Security Check scope, among its review; that review and Aevral's pull-request review, live on install, both read pull requests. See https://aevral.com/compare/greptile. SonarQube runs platform quality and security analysis and its Hunter Agent hunts logic flaws (SonarQube Cloud Enterprise, announced for Server in 2026-08, per Sonar); Aevral reads access control and business logic at rest, and a wider security class list on the pull request. See https://aevral.com/compare/sonarqube. Codex Security is OpenAI's application security agent: find, confirm, and fix vulnerabilities across repository scans and Security Review on pull requests. Aevral competes on the same jobs; a finding can carry a suggested diff you apply plus a fix prompt, and no patches are auto-applied. The dated comparison covers workbench, CLI and SDK, cloud, and preview access. See https://aevral.com/compare/codex-security. ZeroPath publishes business-logic and broken-authentication detection with PR reviews and one-click autofix across a broad AppSec stack; Aevral reads access control and business logic at rest, and a wider security class list on the pull request. See https://aevral.com/compare/zeropath. AISLE runs AI-native vulnerability management in cloud, on-prem, or air-gapped deployments, with business-logic and broken-access-control scope in its AI SAST, fix agents, and agentic verification; Aevral reads access control and business logic at rest, and a wider security class list on the pull request. See https://aevral.com/compare/aisle. Tachyon reviews pull requests in full-codebase context and validates findings for exploitability before reporting, with a fix attached; its published examples include broken access control and IDOR, and both products read pull requests for security. See https://aevral.com/compare/tachyon. Gecko Security analyzes code, logic, and infrastructure with a semantic graph across repos and microservices, with a pull-request review bot and one-click autofix; Aevral reads access control and business logic at rest, and a wider security class list on the pull request. See https://aevral.com/compare/gecko-security. Nullify drives product-security findings to merge-ready remediation PRs with exploit validation, with broken access control and IDOR in its published detection scope; Aevral reads access control and business logic at rest, and a wider security class list on the pull request. See https://aevral.com/compare/nullify.

- Aevral and Claude Security (competitor, even table): https://aevral.com/compare/claude-security
- Aevral and Codex Security (competitor; even 9-row table, dated and sourced): https://aevral.com/compare/codex-security
- Semgrep and Aevral (complementary; overlaps the job: AI-powered detection incl. IDOR and broken authorization): https://aevral.com/compare/semgrep
- Snyk Code and Aevral (complementary): https://aevral.com/compare/snyk-code
- CodeQL and Aevral (complementary): https://aevral.com/compare/codeql
- Aikido and Aevral (complementary; platform plus Code Security Audit): https://aevral.com/compare/aikido
- GitHub Copilot code review and Aevral (complementary): https://aevral.com/compare/github-copilot
- Cursor Bugbot and Aevral (complementary): https://aevral.com/compare/bugbot
- CodeRabbit and Aevral (complementary; reviews plus CodeRabbit Security monitoring): https://aevral.com/compare/coderabbit
- Greptile and Aevral (complementary; review plus Security Check scope): https://aevral.com/compare/greptile
- Qodo and Aevral (complementary; AI code review and governance platform): https://aevral.com/compare/qodo
- Socket and Aevral (complementary): https://aevral.com/compare/socket
- SonarQube and Aevral (complementary; platform plus Hunter Agent): https://aevral.com/compare/sonarqube
- Nullify and Aevral (complementary; autonomous remediation to merge-ready PRs): https://aevral.com/compare/nullify
- Gecko Security and Aevral (complementary; semantic cross-service scanning): https://aevral.com/compare/gecko-security
- Tachyon and Aevral (complementary; exploit-validated PR security reviews): https://aevral.com/compare/tachyon
- AISLE and Aevral (complementary; AI-native vulnerability management, perimeter-deployed): https://aevral.com/compare/aisle
- ZeroPath and Aevral (complementary; AI-native SAST platform plus logic and auth detection): https://aevral.com/compare/zeropath

## Install

Install hub: https://aevral.com/install. One GitHub App plus the console. Install https://github.com/apps/aevral. After the App is installed, the next pull request can already be reviewed, even before anyone signs in.

- GitHub App surface: https://aevral.com/install/github. Install https://github.com/apps/aevral. Scan permissions: Contents read, Metadata, Checks write. PR review adds Pull-requests read and write. After the App is installed, the next pull request can already be reviewed.
- Console surface: https://aevral.com/install/console. Press Scan on the default branch, read findings with evidence, copy a fix prompt. aevr_ keys exist today (Developer API page); billing is console-session-only. No live MCP endpoint today.
- Coding-agent plugin: https://aevral.com/install/agent-plugin. Public repo https://github.com/better-isms/aevral-plugin. Claude: /plugin marketplace add better-isms/aevral-plugin then /plugin install aevral@aevral-plugins. No live MCP endpoint today.

## Who it is for

Personas hub: https://aevral.com/for. One reading, four audiences: PR review, live on install, looks for access control, business logic, injection, XSS, SSRF, path traversal, unsafe deserialization, token and session flaws, and LLM-integration flaws on pull requests, and the scan reads access control and business logic across the repository.

- Founders and org owners who merge their own and their agents' pull requests: https://aevral.com/for/teams-shipping-agent-code
- Open-source maintainers (public repos are the free lane): https://aevral.com/for/open-source-maintainers
- AI security engineers: https://aevral.com/for/ai-security-engineers
- Security engineers: https://aevral.com/for/security-engineers

## Alternatives

Alternatives hub: https://aevral.com/alternatives. Field maps with each tool described by its own stated job and linked to a deeper comparison. Aevral is never listed as a like-for-like alternative to a SAST: on the anchored pages it is what you add alongside: a PR security review that looks for access control, business logic, injection, XSS, SSRF, path traversal, and LLM-integration flaws on the diff, and a whole-repo scan that reads authorization, IDOR, and business-logic access control. The Claude-Security-alternative framing lives on the compare page.

- Whole-repo AI security scanners in 2026 (roundup): https://aevral.com/alternatives/best-whole-repo-ai-scanners
- IDOR and broken access control scanners in 2026 (field map): https://aevral.com/alternatives/idor-and-access-control-scanners
- Semgrep alternatives: https://aevral.com/alternatives/semgrep-alternatives
- Snyk Code alternatives: https://aevral.com/alternatives/snyk-code-alternatives
- Aikido alternatives: https://aevral.com/alternatives/aikido-alternatives
- CodeQL alternatives: https://aevral.com/alternatives/codeql-alternatives
- CodeRabbit alternatives: https://aevral.com/alternatives/coderabbit-alternatives
- SonarQube alternatives: https://aevral.com/alternatives/sonarqube-alternatives
- Greptile alternatives: https://aevral.com/alternatives/greptile-alternatives
- Bugbot alternatives: https://aevral.com/alternatives/bugbot-alternatives
- GitHub Copilot code review alternatives: https://aevral.com/alternatives/github-copilot-alternatives

## Guides

Guides hub: https://aevral.com/guides. Substance pages with primary sources: the access-control surface vocabulary (OWASP API1:2023 / API5:2023, CWE-639, CWE-284), the multi-tenant IDOR deep dive, the pipeline mechanics, the scan-report walkthrough, the fix-prompt walkthrough, the PR access-review checklist, where access control hides in business logic (the third class), the LLM-integration review checklist (OWASP LLM01:2025 / LLM05:2025), per-pull-request pricing explained, IDOR review with a GitHub app, the missing tenant check in a pull request, the Next.js and Supabase authorization check before launch, security review for pull requests written by Claude Code, security review for pull requests written by Cursor, security review for pull requests written by Codex, vibe coding security, the grounded-review noise explainer, and the launch-updates list explainer.

- What a whole-repo authorization scan reads (OWASP API1/API5, CWE-639, CWE-284 cited): https://aevral.com/guides/whole-repo-authorization-scan
- PR security review and SAST are different questions: https://aevral.com/guides/pr-security-review-vs-sast
- Handing a security finding to your coding agent: https://aevral.com/guides/handing-a-fix-prompt
- Reviewing a pull request for access control (the diff review checklist): https://aevral.com/guides/reviewing-a-pr-for-access-control
- Working a scan report of access-control leads: https://aevral.com/guides/working-a-scan-report
- How IDOR happens in multi-tenant code (worked example: the caller-named key, the session tenant, the scope clause): https://aevral.com/guides/idor-in-multi-tenant-code
- Where access control hides in business logic (the third class: entitlements, plan gates, label-matched roles): https://aevral.com/guides/business-logic-access-control
- Reviewing a pull request that wires in an LLM (the LLM-integration checklist, OWASP LLM01:2025 / LLM05:2025 cited): https://aevral.com/guides/reviewing-a-pr-for-llm-integration
- Security review priced per pull request, explained (per-organization allowance, what one review costs on each plan): https://aevral.com/guides/security-review-priced-per-pull-request
- Reviewing pull requests for IDOR with a GitHub app (install to first comment, receipts quoted): https://aevral.com/guides/idor-review-github-app
- Spotting a missing tenant check in a pull request (worked example: a new handler that skips the org-scoped helper): https://aevral.com/guides/missing-tenant-check-in-a-pull-request
- Next.js and Supabase: the authorization check before launch (RLS, secret key, server actions, route handlers): https://aevral.com/guides/nextjs-supabase-authorization-before-launch
- Security review for pull requests written by Claude Code (Claude Code's own review modes, the access-review checklist, and a second reviewer on the repository): https://aevral.com/guides/security-review-for-claude-code-pull-requests
- Security review for pull requests written by Cursor (what Bugbot's docs say it does, and a second reviewer that stays a suggestion): https://aevral.com/guides/security-review-for-cursor-pull-requests
- Security review for pull requests written by Codex (Codex Security's own description, and a read that does not apply the fix): https://aevral.com/guides/security-review-for-codex-pull-requests
- Vibe coding security: who reads the diff (what the word means, why the security question stays, and the three ways the diff gets read): https://aevral.com/guides/vibe-coding-security
- Why review bots get ignored, and what a grounded review does instead (the three habits that get a security bot muted, and the mechanical rules a review needs before it speaks: verbatim grounding on added lines, silence on clean pull requests, unverified leads listed not posted, the cap of five, the record published): https://aevral.com/guides/why-review-bots-get-ignored
- The Aevral launch-updates list, explained: https://aevral.com/guides/waitlist-and-opening

## Vulnerability classes

Vulnerability classes hub: https://aevral.com/vulnerabilities. One page per class that Aevral's PR security review, live on install, looks for on the diff and that has no guide of its own. Each page gives a plain-words explanation, why coding agents write the class, the shapes it takes in a pull request, one illustrated diff, an illustrated finding (a lead with evidence, not a confirmation), the safe version, and what the page does not claim. The whole-repo scan reads authorization, IDOR, and business-logic access control only, not these classes. Access control, IDOR, business logic and LLM integration are covered by the guides above.

- SQL injection in agent-written code (why sort and filter features bring raw SQL back): https://aevral.com/vulnerabilities/sql-injection
- Command injection in agent-written code (shell strings, argument injection, CI steps): https://aevral.com/vulnerabilities/command-injection
- Cross-site scripting in agent-written code (markdown, raw HTML and inline data in React and templates): https://aevral.com/vulnerabilities/cross-site-scripting
- SSRF in agent-written code (link previews, webhooks, import from URL, model fetch tools): https://aevral.com/vulnerabilities/ssrf
- Path traversal in agent-written code (join is not a fence; Zip Slip; upload names): https://aevral.com/vulnerabilities/path-traversal
- Unsafe deserialization in agent-written code (pickle, unsafe YAML loaders, model files): https://aevral.com/vulnerabilities/unsafe-deserialization
- Token and session flaws in agent-written code (JWT decode vs verify, expiry, logout, reset tokens): https://aevral.com/vulnerabilities/token-and-session-flaws

## Glossary

Glossary: https://aevral.com/glossary. One page of anchored, plain-words definitions for the security terms a technical founder meets around AI security review of pull requests: authorization vs authentication, broken access control, business logic flaw, coding agent, cross-site scripting, IDOR, least privilege, path traversal, prompt injection, pull request security review, SAST, security finding, SQL injection, SSRF, tenant isolation, unsafe deserialization. Each term has a one-to-two sentence definition, an everyday example, what it looks like in a pull request, and links to the deeper guide or class page. Terms that fall under a class on the PR review list say Aevral PR review looks for them and name that class; concepts, categories and principles (authorization vs authentication, pull request security review, SAST, least privilege, coding agent, security finding) make no coverage statement. Anchors are the term ids, for example https://aevral.com/glossary#idor.

## Works alongside

Tools Aevral runs beside, each doing its stated job while Aevral adds its PR security review (the general class list on the diff) and its whole-repo access-control scan: https://aevral.com/checks. Each page states the tool's focuses and what Aevral adds alongside it, and links to the fuller comparison. Never a claim about what the other tool would or would not flag.

- Semgrep and Aevral, at a glance (overlaps the job): https://aevral.com/checks/semgrep
- Snyk Code and Aevral, at a glance: https://aevral.com/checks/snyk-code
- CodeQL and Aevral, at a glance: https://aevral.com/checks/codeql
- Aikido and Aevral, at a glance (platform plus Code Security Audit): https://aevral.com/checks/aikido
- GitHub Copilot code review and Aevral, at a glance: https://aevral.com/checks/github-copilot
- Cursor Bugbot and Aevral, at a glance: https://aevral.com/checks/bugbot
- CodeRabbit and Aevral, at a glance (reviews plus CodeRabbit Security monitoring): https://aevral.com/checks/coderabbit
- Greptile and Aevral, at a glance (review plus Security Check scope): https://aevral.com/checks/greptile
- Qodo and Aevral, at a glance (code review and governance): https://aevral.com/checks/qodo
- Socket and Aevral, at a glance: https://aevral.com/checks/socket
- SonarQube and Aevral, at a glance (platform plus Hunter Agent): https://aevral.com/checks/sonarqube
- Nullify and Aevral, at a glance (remediation agents): https://aevral.com/checks/nullify
- Gecko Security and Aevral, at a glance (semantic scanning): https://aevral.com/checks/gecko-security
- Tachyon and Aevral, at a glance (PR security reviews): https://aevral.com/checks/tachyon
- AISLE and Aevral, at a glance (vulnerability management): https://aevral.com/checks/aisle
- ZeroPath and Aevral, at a glance (AI-native SAST platform): https://aevral.com/checks/zeropath

Alongside the platforms startups use for SOC 2 or ISO 27001. Each page covers what the platform documents about GitHub pull requests (branch rules, approval counts, or the review record); Aevral reads what the pull request changes, for security. Each page quotes the platform's own documentation with the retrieval date.

- Vanta and Aevral (its Application Changes Reviewed test, plus Aevral's security reading of the diff; no native integration): https://aevral.com/checks/vanta
- Drata and Aevral (its Test 8, Formal Code Review Process, plus Aevral's security reading of the diff; no native integration): https://aevral.com/checks/drata
- Secureframe and Aevral (the required approval counts it reads, plus Aevral's security reading of the diff; no native integration): https://aevral.com/checks/secureframe
- Sprinto and Aevral (its peer review check, plus Aevral's security reading of the diff; no native integration): https://aevral.com/checks/sprinto
- Hyperproof and Aevral (the pull request review record it collects, plus Aevral's security reading of the diff; no native integration): https://aevral.com/checks/hyperproof
- Scrut and Aevral (its Code Review Results task, plus Aevral's security reading of the diff; no native integration): https://aevral.com/checks/scrut

## Company

Aevral is an AI security reviewer for GitHub pull requests. PR security review (live on install) looks for access control, business logic, injection, XSS, SSRF, path traversal, unsafe deserialization, token and session flaws, and LLM-integration security. Whole-repo scans read access control and business logic, with a suggested fix you apply in Claude Code, Cursor, or Codex. Built by Better ISMS, the company behind ISMS Copilot, in Paris, France. Better ISMS EURL is a one-founder software company run by Tristan Roth. The same company builds ISMS Copilot, the ISO 27001 and GRC assistant, and heyGRC, the pull-request compliance reviewer; ismsdirectory.com is the family directory. One EURL, no outside investors. Better ISMS is getting ISO 27001 certified and is not certified today; the intended scope includes Aevral. Key facts: launched 2026; registered address 60 rue François 1er, 75008 Paris, France; pricing per organization, not per seat (PR free tier live: 500 public reviews a month, 25 private a month after the 14-day trial; paid PR plans from $49 per month; scan plans from USD 0 for public repos to USD 1,699 self-serve; Enterprise quoted from USD 3,300 per month equivalent; HT, B2B, excluding VAT); self-serve plans are monthly, Enterprise is an annual commitment with invoice billing; no customer names are published; named competitors Claude Security (Anthropic) and Codex Security (OpenAI), with Aikido, CodeRabbit, Greptile, SonarQube, and others complementary (https://aevral.com/compare). Founder socials: X https://x.com/Tristanyway and GitHub https://github.com/tristanroth1. About page: https://aevral.com/about. Contact form: https://tally.so/r/PdxdbV. Aevral itself is not open source; the models are.

## Pages

Home https://aevral.com. Products hub https://aevral.com/products. Scan product https://aevral.com/products/scan. PR review product https://aevral.com/products/pr-review. Pricing (both products) https://aevral.com/pricing. Compare hub https://aevral.com/compare, last verified 2026-09-29 against https://claude.com/product/claude-security and https://support.claude.com/en/articles/14661296-use-claude-security for the Claude Security rows. Works-alongside hub https://aevral.com/checks. FAQ https://aevral.com/faq. About https://aevral.com/about. Manifesto (why Aevral exists, the evergreen argument: agents write more code than anyone reads, so security review has to run on its own, survive switching agents, and show its evidence) https://aevral.com/manifesto. For agents https://aevral.com/for-agents. Trust: https://trust.ismscopilot.com. ISO 27001 programme (not certified today; intended company scope includes Aevral): https://trust.ismscopilot.com/en/iso-27001. Privacy: https://trust.ismscopilot.com/privacy-policy. Legal entity: Better ISMS (Better ISMS EURL), 60 rue François 1er, 75008 Paris. Security (what the GitHub App can do, what code is read and where it goes, storage and retention, model path, gaps and announced changes; the DPA section 1.7 and the Aevral sub-processor page https://trust.ismscopilot.com/en/aevral/subprocessors are binding) https://aevral.com/security. Vulnerability disclosure policy https://aevral.com/security/disclosure. Transparency report (scope, method, full frozen eval matrix, corpus digest, reproduction protocol) https://aevral.com/transparency. Receipts (the named PR review eval runs of 2026-09-25 and 2026-10-01: corpus, recall per run, fabricated findings, the production configuration, and the stated limits; new runs of the customer setup get appended as they happen, worse results included) https://aevral.com/receipts. Audit evidence (what to show an ISO 27001 auditor when you use Aevral: what the review checks for, the secure-development controls the review trail supports as evidence, and where the record and the conclusion live on the pull request) https://aevral.com/audit-evidence. US buyers (what a US purchase process needs from a French vendor: the DPA with its United States State Privacy Laws section, the sub-processor list, the California privacy position, and the honest certification state, each linked to its binding source) https://aevral.com/us-buyers. Aevral is not affiliated with Anthropic.

## Setup and ongoing coverage
Installing the App starts PR reviews, before anyone signs in (25 free private reviews a month). An opt-out stays off while the App stays installed; reinstalling the App turns reviews back on. Setup Complete can turn reviews off. You press Scan; Setup does not start a scan for you. Free organizations get two private scans in 14 days; paid organizations use included scan quota. No setup overage.
Team USD 99 / 4, Business USD 399 / 16 and Scale USD 1,699 / 100 share their included allowance between manual and recurring scans. Each configured repository gets up to four checkpoints per billing month (start plus 0/7/14/21 days), not permanent free weekly scans. Unchanged commits reuse the existing result within the billing period; quota exhaustion waits for renewal. Automatic scans never incur overage.
PR plans: Free 25 private reviews per calendar month after the existing 14-day/500-review trial; Starter USD 49 / 100 PR reviews, Pro USD 199 / 500, Business USD 999 / 3,000, Scale USD 1,999 / 6,500 (one PR review per pull request per billing period, re-pushes do not count again). After a one-time first-overrun buffer (an extra quarter of the plan, free), paid extra PR reviews cost USD 0.99 each only if the owner turned extras on, at checkout or later in the console. 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 (not live yet). Scale carries everything in Business. Enterprise from USD 4,000 a month, not self-serve. Scan and PR review allowances are separate.