Skip to content
Sebastion AI · by Foundation Machines

An autonomous security reviewer that lives on your pull requests.

Install once. Sebastion runs three scanner classes on every PR — stack-aware AI SAST, regex-based secret detection, and OSV.dev SCA — and posts findings as inline review comments tagged with severity and CWE id, with a concrete fix. Free for public repos.

github.com/acme/agents · issue #128live
Sebastion AI · opened just now

Sebastion AI security audit: PR #42

Cloning repository at PR head…

3
Scanner classes per audit
12
Stack-aware framework profiles
Inline
Review comments, line-level
Free
On every public repo
What Sebastion does

Three scanners. One issue. One concrete fix.

Sebastion is not a model — it is an agent. It composes deterministic scanners (OSV-Scanner) with frontier-model reasoning (Claude) to turn diffs into actionable security findings.

SAST

Stack-aware vulnerability reasoning.

Sebastion knows your stack — Django, FastAPI, Flask, Next, Express, Go, Spring, Rust — and applies the framework-specific vulnerability checklist on every PR. Catches injection, broken auth, insecure crypto, unsafe deserialisation, SSRF, path traversal and the long tail of OWASP Top 10 issues with the diff as context.

Secrets

Verified credentials, before they merge.

Regex-based detection for AWS access keys, GitHub PATs (classic + fine-grained), OpenAI / Anthropic / Stripe / Slack / Google API keys, JWTs (with header validation), and private-key blocks. Diff-additions only, with EXAMPLE / placeholder filtering to keep the false-positive rate low.

SCA

OSV-Scanner on every dep manifest change.

When `package.json`, `requirements.txt`, `go.mod`, `Cargo.toml`, `Gemfile.lock` or any supported manifest changes, Sebastion queries OSV.dev for known CVEs against the new versions. Real CVE ids, NVD advisory links, and concrete fix versions land as findings on the same PR.

Output

Inline PR review comments, CWE-tagged.

Findings post as line-level review comments on the offending change, tagged with severity emoji and a MITRE CWE link. Critical findings flip the review to CHANGES_REQUESTED so the PR is blocked from auto-merge. No new dashboard to learn, no email digest, no separate scanner repo.

How it works

From PR open to filed finding in under a minute.

Same pipeline on every push. Re-runs incrementally when you push a fix, so the issue body stays in sync with HEAD.

Step 01

Install on GitHub

One click on github.com/apps/sebastionai. Sebastion gets scoped, short-lived read access to the repos you choose. No PATs, no SSH keys, no seat invitations.

Step 02

Audit on every PR open or push

On every PR event, Sebastion fetches the diff via the GitHub API and routes it through the three scanners in parallel. No code clone, no source persisted, ephemeral execution.

Step 03

Run three scanners in parallel

Stack-aware AI audit (Claude Sonnet 4.6 on Free, Opus on Pro+) for SAST. OSV.dev for SCA on any manifest changes. Regex-based secret detection over diff additions. Findings are merged, deduplicated and severity-sorted.

Step 04

Post inline review + flip merge state

Findings land as inline review comments on the offending lines, each with severity, rule id, CWE link, and a concrete fix. Critical findings set the review to CHANGES_REQUESTED. Re-runs on every push, never blocks merge if no critical findings exist.

The models behind Sebastion

Pick a tier, not a model.

Each tier wires Sebastion to the right model for the job. Routed via OpenRouter so we can swap the underlying provider without changing your install.

Free
$0forever

Public repos, stack-aware AI, OSV CVEs + secret scan.

Powered by Claude Sonnet 4.6.

Install on GitHub
Coming soonPro
$19/ developer / month

Private repos, Opus reasoning, autofix PR drafts.

Powered by Claude Opus 4.7.

Coming soonTeam
$39/ developer / month

Multi-org, custom rules, multi-model cross-check.

Opus 4.7 + cross-check across vendors.

Enterprise
Custom

Self-host runner inside your VPC, SSO, SOC 2, SLA.

Best frontier models on request.

Talk to us

Full feature comparison and per-developer billing details on the pricing page.

Who uses Sebastion

Three audiences. One install.

OSS maintainers

Free, in-issue, no service to run.

Sebastion installs in a click and reviews every PR from a contributor. No new build step, no separate scanner repo, no badge fatigue. Public repos are free forever.

AI startups

Catch the security bugs before merge.

AI codebases ship fast and the model writes a lot of the diff. Sebastion is the second pair of eyes that doesn't get tired — calling out the SQL injection, the leaked key, the vulnerable dep, before anyone hits squash and merge.

Regulated workloads

Self-host, audit log, SOC 2.

Enterprise tier runs the audit runner inside your VPC, with SSO/SAML, an audit log of every Sebastion action, a signed DPA and SOC 2 attestation. The same auditor, on infrastructure you control.

Privacy by construction

Your code is handled, not stored.

Sebastion never persists your source. PR diffs are fetched via the GitHub API, routed through scanners, and discarded as soon as the audit completes. We never train models on your data and no-train flags are set on every upstream request that supports them.

  • No persistent storage of source code. PR diffs and file contents live only for the duration of one audit run.
  • Findings are owned by you — they live in your repo as inline PR review comments.
  • LLM calls are routed through OpenRouter to upstream providers (Anthropic, OpenAI, Google) with no-train flags set.
  • Read the privacy policy and security page for the full surface.

Catch the security bugs before merge.

Install Sebastion on your GitHub org and review your next pull request in under a minute. Free for every public repo.