Existing checks
Build, types, lint and tests.
Open source · An earlier signal
See the issue Verify the fix
Run checks. Locate failures. Verify repairs.
FOLLOW ONE FAILUREInteractive example
Connected by context
Interface01
Application02
Evidence03
One check needs attentioncart.test.ts → checkout.ts
ASSERTION FAILEDcart.test.ts:18
Expected 100. Received 110. The discount was never applied.
24
return subtotal + shipping;
One entry point for your project checks
Node.jsTypeScriptPythonGoRust
THE CONNECTED WORKFLOW
From first signal to verified fix
Code in context
Architecture, source and coverage.
Traceable evidence
Original errors, hashes and history.
Verified reruns
Before, after and what changed.
Your project, in focus
Checks, architecture and evidence in one workspace
canary / workspace
Actual product interfaceReports: JSON · JUnit · Markdown
01
Install once
curl -fsSL https://raw.githubusercontent.com/EVEDensity/Canary/main/scripts/install/install.sh |
bash
02
New terminal. Enter your project. Run checks.
canary run --ci
03
Or run with the interactive report
canary run --port 4318
Node.js 24 & Git. Project dependencies ready.Installation guide
Do I need to change my existing CI?+
Keep your existing workflows. Canary runs project checks through a CLI with stable exit codes and standard reports. Use it before a commit or invoke it in your current CI pipeline.
Does running checks require an AI model?+
No model is required for automatic project checks. Agent evaluations and model-based evaluators use the adapters and credentials you explicitly configure.
What does automatic discovery actually run?+
Canary reads existing Node scripts and standard Python, Go and Rust test entry points. Node/Python automatic execution has recorded verification. It runs your existing checks; it does not create tests or automatically collect coverage for every project. Existing Canary configuration takes priority.
Where do my reports live?+
Run artifacts are stored in your project's .canary directory. Manifests, content hashes and lineage help connect results to the run that produced them. Review and redact evidence before sharing it.