dagsec reads every commit for leaked credentials and scores every dependency your code pulls in, then reports on the pull request before it merges.
Runs inside your CI. Your code never leaves your runner.
A pull request, a merge request or an AI agent's npm install starts the check.
On your CI runner it reads the full git history for secrets; for every package it checks known vulnerabilities and a 0–100 health score.
A single comment on the pull request with a pass or fail check, or the install stopped with the safe version named.
A pull request with a leaked key, a dashboard scan with an SBOM, and an AI agent stopped before it installs a vulnerable package.
When a pull request adds a credential or depends on a package that is falling apart, dagsec says so in the pull request itself and marks the check as failed. The comment updates on every push instead of piling up.
AKIA********, never the secret..dagsecignore, read from the main branch so a pull request can't silence itself.dagsec scan
Failed: 1 secret, 0 of 14 dependencies below 40
| Secret | File | Value |
|---|---|---|
| AWS access key ID | config/deploy.env:3 | AKIA******** |
Rotate this credential: removing it from the code does not remove it from git history.
A 0 to 100 score from six signals, taken from the package registry, GitHub and OSV.dev. When a signal can't be measured it is left out of the score, not guessed. These two were measured in September 2026.
requests95Maintained, widely used, three people wrote half the code.
left-pad58Still downloaded 1.8M times a week from an archived, single-author repository.
Coding agents pick package versions from training data that can be a year old, and sometimes invent package names that don't exist. dagsec checks every package against today's vulnerability data before the install runs.
Install in Claude Code:
/plugin marketplace add hasanerman/dagsec-claude
/plugin install dagsec@dagsecIncluded in every plan, Free too. Only package names and install commands are sent, never your code.
Add lodash 4.17.4 to this project
$ npm install lodash@4.17.4dagsec blocked this install:
- lodash@4.17.4 has 4 critical/high vulnerabilities (CVE-2019-10744, CVE-2021-23337, …). Install lodash@4.18.0 or later instead.
That version has known critical vulnerabilities, so I installed lodash@4.18.1 instead. The API is the same.
Everything on app.dagsec.net, plus the parts that only work on your own computer. Sign in with GitHub and the app finds your AI coding tools and protects them in one click.
dagsec-setup.exe. If SmartScreen appears, choose More info and Run anyway: the app isn't code-signed yet.dagsec.dmg and drag dagsec to Applications. The first time, right-click it and choose Open, or allow it under System Settings → Privacy & Security.dagsec command and checks the installs you type in the terminal.Windows 10 and 11 (64-bit), macOS 11 or later (Apple Silicon and Intel). Removing the app takes out everything it added to your tools and repositories.
DAGSEC_API_KEY (Settings → Secrets and variables → Actions)..github/workflows/dagsec.yml to the repository and open a pull request. dagsec comments with its report and marks the check.npm, PyPI, crates.io, Go, Maven, Gradle and NuGet, across every manifest in a monorepo. Public repositories are free; private repositories need a paid plan.
name: dagsec
on: pull_request
permissions:
contents: read
pull-requests: write
id-token: write
jobs:
dagsec:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
with:
fetch-depth: 0
- uses: hasanerman/dagsec-action@v1
with:
api-key: ${{ secrets.DAGSEC_API_KEY }}
fail-under: 40Secret scanning, dependency auditing and AI agent guardrails usually come from separate products, each with its own setup and bill.
| Usually | dagsec | |
|---|---|---|
| Leaked secrets across the full git history | A secret scanner | Included |
| Known vulnerabilities in pinned versions | A dependency (SCA) tool | Included |
| Health score for abandoned or single-maintainer packages | Manual review | Included |
| Stop an AI agent's risky install before it runs | Not covered | Included |
| SBOM (CycloneDX) and license policy | A separate compliance tool | Included |
| Runs on your CI runner; code never uploaded | Often a cloud upload | Included |
No. The GitHub Action and the GitLab job run on your own CI runner. dagsec's server only hands out the scanner and receives the result's counts. Dashboard scans clone public repositories, or private ones you connect through the GitHub App, and delete the clone right after.
Only the install command and the package names and versions, for example npm install lodash@4.17.4. File contents and other commands never leave your machine.
Your work continues. When dagsec can't be reached, CI and agent checks let things through instead of blocking you, and the failure is logged.
npm, PyPI, crates.io, Go, Maven, Gradle and NuGet for vulnerabilities; npm, PyPI and crates.io also get the health score. Secret scanning works on any repository.
Yes, on the Student, Pro and Team plans. The Free plan covers public repositories, verified through GitHub's or GitLab's signed CI identity.
On a dedicated server at Hetzner in Germany (EU). Security details and the data processing agreement are on the Security page.
Start free on public repositories. Upgrade when you need private repositories or more runs.
$0 / month
For open-source projects and trying dagsec out.
$11 / month
For students with a school email address.
$25 / month
For developers and small teams with private code.
$45 / month
For companies running dagsec across many repositories.
Prices in US dollars. Tax is added at checkout where it applies. Cancel any time; 14-day money-back guarantee.
Sign in with GitHub and scan any public repository. No card needed.
Start free