Discover vulnerabilities, trace real attack paths across files, detect exposed secrets and govern AI-generated code.
Measured on 81,324 labelled benchmark samples: 80% of vulnerabilities found, 1.0% of safe samples flagged. See the full report
Add CSV export to billing
#482 · feature/billing-export · 14 files
Risk · High
74
62%
AI-generated
2
Vulnerabilities traced
1
Secrets
3
Reachable deps
src/billing/export.ts
SQL injection traced across 2 files
src/auth/session.ts
78% AI-generated · Cursor
src/ui/Table.tsx
No findings
Merge held until the 2 flagged files are signed off by a named reviewer.
An example of one pull request check · scan a sample app yourself
0
language engines
0
data-flow rules
0
CWEs mapped
0
AI-detection signals
0
IaC & cloud formats
0
API languages inventoried
TrustLedger is in early access. Teams that join now use the full product for free while we shape it with their feedback.
What you receive
How to start
Questions first? Contact the team or read the getting started guide.
Six areas, one scan: pick a tab to see what each one covers.
A single organization-wide score blends attestation coverage, policy compliance, and critical-risk resolution into one number that trends over time — the same view your team and your board can both read.
What you will find in the sidebar once you connect a repository.
Overview
3 pagesOverview
Org health score, per-repo breakdown and what needs attention now.
Analytics
Scan volume, finding trends and time-to-attest.
TrustScore™
One blended score across attestation, policy and risk resolution.
Compliance
3 pagesCompliance
Evidence mapped to SOC 2, the EU AI Act and PCI-DSS.
SLA Dashboard
What is on track, at risk or overdue.
Risk Register
CWE-mapped risks with assigned reviewers and deadlines.
Code Risk
4 pagesScan History
Every scan, with per-file risk and source-to-sink traces.
Secrets
Hardcoded keys and tokens, caught as they land.
Dependencies
Live CVEs ranked by whether your code actually reaches them.
API Security
Every endpoint inventoried, and the route missing the auth its siblings have.
AI Intel
1 pageShadow AI
AI tools that are not on your approved list, confidence-scored.
Audit
4 pagesAudit Trail
An immutable record of every attestation, scan and override.
Reports
Signed evidence packages and SBOM exports (SPDX, CycloneDX).
Trace
Follow any scan step by step to see why a result looks the way it does.
Rule Catalog
Every rule with its CWE, what it flags and how to fix it.
Threats
1 pageThreats
Policy breaches, alerts and incidents with SLA deadlines, plus threat intelligence on patterns found in your repos.
No CI/CD changes. No config files. Install once and every PR is scanned automatically.
Install the GitHub App in under 2 minutes (GitLab and Bitbucket also supported). Every pull request is scanned automatically — no config files, no CI changes.
AI% per file, secrets, reachable vulnerable dependencies, traced vulnerabilities, and cloud, container and API misconfigurations — each with real evidence, not a guess. One risk score ties it together.
Reviewers sign off on flagged files directly in the dashboard. Policy gates block merges until required attestations and fixes are recorded.
Prefer to drive it yourself? Scans can also be submitted through the REST API.
Every finding comes with the path an attacker's value takes to the dangerous call — so reviewers spend their time on judgment calls, not on re-deriving whether it's real.
req.query.host
const cmd = `ping -c 1 ${host}`
if (debugMode) {…} else {…} — taint survives both arms
crosses into ./shell via "runPing"
exec(cmd)
stable ID assigned — survives unrelated edits to this file
vs Point tools (one per risk type)
A SAST tool for vulnerabilities, a separate secrets scanner, an SCA tool, an IaC scanner, an API inventory, and no visibility into what's actually AI-generated — a dashboard and an audit trail for each, nothing reconciled.
One platform: vulnerabilities, secrets, dependencies, cloud and container config, API auth coverage, and AI provenance scored together on every PR, with one policy engine and one audit trail behind all of it.
vs SCA that lists every CVE
Every CVE in every declared package, sorted by CVSS — including packages nothing imports and functions nothing calls. Teams learn to ignore the list.
Each vulnerable package is ranked by reachability: the vulnerable function called from a route, imported from reachable code, imported only by tests, or never imported — with the file and line as evidence.
vs Regex-based scanners
Match a pattern on one line. No concept of whether the value is actually attacker-controlled, whether it was sanitized first, or whether it even reaches a sink.
Real data-flow: a bitmask of sink classes carried from source to sink, cleared only by a sanitizer that actually neutralizes that class.
vs Manual review + spreadsheets
Reviewers eyeball the diff, chase compliance evidence by hand, and track AI tool usage — if at all — in a doc that's out of date by Friday.
Every PR gets a risk score, a named attestation, and an audit-ready trail. Shadow AI usage, policy exceptions, and SLA breaches live in one system that's always current.
No. TrustLedger installs as a GitHub App — GitLab and Bitbucket are supported too — and scans every pull request automatically, with no config files and no CI changes. If you'd rather drive it yourself, scans can also be submitted through the REST API.
Ten languages have their own AST parser and taint engine: TypeScript/JavaScript, Python, Java, Kotlin, Go, C#, PHP, Ruby, Rust and C/C++ (which also checks memory safety: use-after-free, double free, unchecked allocations, buffer overflows), tuned to frameworks such as Express, NestJS, Flask, FastAPI, Spring, Ktor, ASP.NET, Laravel, Rails, Axum, Actix and Rocket. Infrastructure checks cover Terraform, CloudFormation, ARM, Bicep, Serverless, Kubernetes, Helm, Docker and OpenAPI.
A vulnerability is reported only when an attacker-controlled value is traced to a real sink — through assignments, branches, helper functions and other files — and a sanitizer clears only the vulnerability classes it actually neutralizes. Every finding ships with its source-to-sink trace, so reviewers can check it in seconds.
47 signals across code structure, git history and behavior estimate how much of each file is AI-generated and attribute it to the tool that wrote it — Copilot, Cursor, Claude, Windsurf and more — and flag tools that aren't on your approved list.
Yes. Policy gates hold a pull request until the required fixes or a named reviewer's attestation are recorded, and the result is posted back to the PR as a check run.
Attestations, exceptions and findings land in an immutable audit trail and export as signed evidence packages mapped to SOC 2, the EU AI Act and PCI-DSS, along with SBOMs in SPDX and CycloneDX.

AI-generated code, leaked secrets, reachable vulnerable dependencies, a public bucket, an endpoint that forgot its auth check — any of it can slip into a PR unnoticed. TrustLedger scores it, traces it, and makes sure a human signed off before any of it reaches production.
Connecting asks for read access to pull requests and code, and write access to checks — why each permission. Free to start, no credit card.
Questions or pricing? Talk to us — trustledgerapp.dev@gmail.com