Get started
Set up Stryct in your CI on a GitHub repo. Results show up in your pipeline and in Stryct Studio.
01Create an accountAvailable now
Sign up at stryct.ai/signup with email, Google or GitHub. This gives you a Stryct Studio account.
02Get an API keyComing soon
In Stryct Studio, open Settings → API keys and create a key. Copy it right away. It is shown once.
Use one key per repo, so you can revoke one without touching the others.
03Add the key to your repoComing soon
In GitHub, open your repo’s Settings → Secrets and variables → Actions and add a secret named TAINT_API_KEY. Or use the GitHub CLI:
gh secret set TAINT_API_KEY04Add the GitHub ActionComing soon
Create .github/workflows/taint.yml:
name: Stryct
on:
push:
branches: [main]
release:
types: [published]
workflow_dispatch:
permissions:
contents: read
checks: write
jobs:
taint:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: Taint-AI/taint-action@v1
with:
api-key: ${{ secrets.TAINT_API_KEY }}
fail-on: highThis runs Stryct when you merge to main, when you publish a release, and when you start it by hand from the Actions tab. fail-on sets the lowest severity that fails the check: critical, high, medium or low.
05Run itComing soon
On each run, Stryct:
- Discovers your agents, models, prompts, tools, MCP servers and auth from the repo.
- Compares them with the last run, so only what changed is tested again.
- Generates attacks for what it found.
- Runs them in a sandbox, each next to a normal version of the same task.
- Reports what broke and suggests a fix.
06Run on pull requests (optional)Coming soon
A full evaluation on every pull request costs more than most teams need. To run it only on pull requests that change your agent, add a pull_request trigger with a paths filter:
on:
pull_request:
paths:
- "agents/**"
- "prompts/**"
- ".mcp.json"Add pull-requests: write to permissions so Stryct can comment on the pull request. Change the paths to wherever your agent code, prompts and MCP config live.
07Read the resultsComing soon
Results show up in three places:
- On the commit or release. A Stryct check that passes or fails. If you turned on pull request runs, one comment on the pull request with the findings and a suggested fix for each.
- In the pipeline. A summary in the job log. The job fails when a finding is at or above
fail-on, so the rest of your pipeline can depend on it. - In Stryct Studio. Every run is uploaded to your dashboard: findings, evidence, suggested fixes and history for each repo. Security teams see coverage, risk changes and exceptions across repos.
08Use it as a release gateComing soon
Make your deploy job depend on the Stryct job (needs: taint), so a release that fails the evaluation does not ship.
If you turned on pull request runs, you can also mark the Stryct check as required in Settings → Branches. A pull request that fails it cannot be merged.