Repo understanding without model tokens.
Codna parses symbols, imports, call paths, tests, and dependencies into a graph the agent can query.
Symbols mappedevery symbol
Bundle sizeprinted on every run
Model spend to map0 tokens
Product
The coding agent with a map.
Codna maps your repo for zero tokens. Then an agent ships a fix with root cause, confidence and regression risk attached, for about $0.02 a fix on average.
Product
The agent receives evidence, not the whole repo.
Instead of dumping files into a context window, Codna builds a compact bundle: suspect files, call chain, failing test, and risk map. The report states what it found.
✓ codna fixed . root cause : <one sentence> symbol : <function> (blast radius: N) confidence : N% · regression risk: N% context : raw → bundle tokens (N× smaller)
Core capabilities
Understand, fix, review, secure. Every result explains itself.
1
Triage
Map any local path or git URL for zero model tokens and see where the change belongs.
2
Fix
Fix from an evidence bundle scoped to the issue. Root cause, confidence and regression risk attached. Add --open-pr to open the pull request.
3
Review
Review every pull request. Findings carry severity, category and confidence. A clean diff gets an approval.
4
Secure
Prove which scanner findings are reachable. Codna reads SARIF from CodeQL, Semgrep, Snyk or Trivy and spends zero model tokens.
Distribution
Four ways to run Codna.
CLIRun codna triage, fix, review and secure in any repo, CI job or container.
MCP serverGive Cursor and Claude five Codna tools: triage, fix, secure, recall and report.
GitHub App and ActionReview every pull request. Open fix PRs from an issue label, a comment or a red check.
Frequently asked
Codna builds a dependency and blast-radius graph of the repository without a model, for zero tokens. From that graph it packs an evidence bundle scoped to the issue and prints the raw-to-bundle size on every run. The agent works only on what matters.
Every fix reports root cause, confidence, blast radius and regression risk, and passes a risk gate before it is applied or a pull request opens. Run codna fix --tests --apply and Codna runs your tests in a sandbox and re-fixes until they pass.
The repository map is built from import patterns, so it is not tied to one language. On-device recall covers Python, JavaScript, TypeScript, Go, Rust, Java, C, C++, C#, PHP and Ruby. Set fix.test_command in codna.yaml to your own test runner.
On 87 matched bug-fix cases against Cursor, Codna used 5× fewer tokens and ran 1.7× faster. Both verified 87 of 87 fixes. Codna averaged about $0.02 of model spend per fix.
Codna ships as a CLI, an MCP server for Cursor and Claude, a GitHub Action, and a GitHub App that reviews pull requests and opens fix PRs in your repo.
Understanding runs on your machine. Only the evidence bundle or the diff reaches your model provider, with your key. Your code is not used to train models unless you opt in.
