The problem
Security teams drown in dependency findings that nobody owns. In our census of 20 popular open-source projects (lockfiles only, not a random sample):
- 248 findings across 11,244 dependencies.
- 95.2% of those findings were noise (not actively exploited, low chance of exploitation), so 3 monsters were spawned. Another 9 findings are unresolved and stay in the report as findings, not as clean.
- 84.7% of dependencies had no CODEOWNERS owner, and 16 of 20 repositories have no CODEOWNERS file at all.
Source: docs/census/RESULTS.md, scan date 2026-11-17. Aggregates only, no project is named. Reproduce with npm run census.
The process
- DetectOSV, CISA KEV, EPSS
- Triagenoise line
- AssignCODEOWNERS
- Fixone PR
- VerifyCI and inquisitor
- Approvea human merges
- Healreraise the plate
- ReportCISO page, OCSF, SLA
Source: docs/ROADMAP.md, section 3. CRYPT opens the pull request and a human merges it.
A two-week read-only pilot
One repository at a time, two weeks. CRYPT starts read-only: it reads, it ranks, it writes reports. It writes nothing to your repositories until you say so.
- Days 1 to 2. Run
crypt graveyard --org <your-org> --repos <owner/repo> --token-env <VAR>on the repositories you chose, with a read-only token. It writes a local file for the Graveyard page with the noise line applied, and nothing leaves your machine. - Days 3 to 5. Walk through the War Room and the CISO report on your own data. Decide which repositories get fix pull requests.
- Week 2. Enable the GitHub Action in its default report mode, which is read-only. For fix pull requests, run a live raid with a fine-grained token for one repository; CRYPT opens at most 5 pull requests per raid. Your team reviews and merges. Automatic pull requests from the Action on your own repositories are planned, not shipped.
- End of week 2. You decide: stop, or continue. Revoke the tokens and remove the workflow file from each repository. Nothing remains running.
Source: docs/ROADMAP.md, P5 and token scopes. No customer has completed this pilot yet.
Time saved ESTIMATE
hours saved per month = repos × findings per repo per month × noise share × triage minutes ÷ 60
An estimate, not a measurement. Change the inputs to match your team. Everything is computed in your browser and nothing is sent.
ESTIMATE: 200 findings, 190 filtered as noise, about 48 hours saved per month.
Defaults: 20 repositories, 10 findings per repo (median before filtering) and 95.2% noise are from the census. The census median is per scan, and we use it as a per-month figure here, which is an assumption. 15 triage minutes per finding is our assumption, not measured, and it counts every filtered finding as a full triage slot saved, which is an upper bound. Replace it with your own.
Pricing HYPOTHESIS
No customers, no revenue and no price exist today. Nothing below has been tested with a buyer.
- Free: public repositories, the CLI and the Action.
- Per-organization tier (HYPOTHESIS): private repositories and the org views.
- Enterprise (HYPOTHESIS): SSO, audit retention, SIEM export.
Source and competitor pricing: docs/BUSINESS.md.
Security review
What data CRYPT reads, stores and sends, which token scopes it needs, and what it does not offer: docs/SECURITY-REVIEW.md.
Start
The issue is public. Share your role, team size and tool names only.