Today's coding agents brute-force your repo.
They read file after file inside the model — burning tokens, time, and context — just to figure out where the bug is. It is slow, expensive, and it forgets.
100Ktokens just to orient.
codna · Understand. Fix. Evolve.
Agents read your code. Codna understands it.
Codna maps your entire repo in milliseconds — for zero tokens — then sends an agent in with exactly the context it needs.
pip install codnaRuns on your machine or your cloud. Your code never leaves without your key.
$ codna fix . --issue "checkout test failing"
✓mapped 1,284 files · 0 tokens · 61ms
→Telys recall: 4 files · 0 cloud round-trips for recall
→root cause: session race · checkout.ts:42
✓patch ready · confidence 91% · risk low
✓test passed · opening pull request
$
5×fewer tokens vs Cursor
1.7×faster than Cursor
$0.02per verified fix
0tokens to understand
Measured across 1,000+ verified tasks — 5× fewer tokens and 1.7× faster than Cursor, 100% verified (vs Cline's 73.6%).
Codna understands first, then acts.
A deterministic engine maps every symbol, dependency, and blast radius — for zero tokens — then hands the agent a tight evidence bundle: 162× less context than reading the repo. And Codna remembers: recall runs on Telys, our custom on-device memory engine.
~600tokens of focused context.
Memory
Codna doesn't forget your codebase.
A deterministic graph engine maps structure for zero tokens; Telys, our custom on-device memory engine, recalls the relevant code on-device. Together they hand the agent a smaller, faster evidence bundle — and the memory persists across sessions.
Proof: structure + recall, measured
Deterministic graph for structure. Telys for on-device recall. Aggressive context compression, with 1,000+ verified tasks.
162×less context than the repo
5×fewer tokens per fix
1,000+verified tasks
0cloud round-trips for recall
How it works
One pass, start to finish.
The agent does not wander through your repo. Codna gives it a map, evidence, and a safe path to ship.
Understand
Codna parses your repo into a dependency and blast-radius graph — no LLM, about 60ms per repo, zero tokens.
Fix
It pinpoints suspect files, packs a minimal evidence bundle, and runs the agent surgically with root cause and confidence.
Evolve
Every fix is simulated for blast radius and risk before you apply it. Codna gets sharper the more you ship.
Benchmarks
Faster. Cheaper. Verified.
Head-to-head on real public repos — single bugs, multiple bugs in one repo, and repos the agent had never seen. Codna used far fewer tokens and ran faster, and every fix passed the test.
| Codna | Codex CLI | Gemini CLI | |
|---|---|---|---|
| Tokens to fix · 8 scenarios | 465K | 6.0M | 4.5M |
| Time | 108s | 568s | 1,000s |
| Fixes verified | 8 / 8 | 8 / 8 | 8 / 8 |
99.9%of the code running on GitHub today — 250+ languages, understood for 0 tokens.
162× lesscontext handed to the agent vs. reading the whole repo.
~$0.04per verified fix — about 28× cheaper than a typical agent edit.
Every scenario verified by a passing test. Full methodology and per-scenario data on the benchmarks page.
250+ languages · zero tokens
Codna maps repositories in 250+ programming languages — Python, TypeScript, Go, Rust, Java and more — for zero tokens.
Workflow
Built into your workflow.
One command, one GitHub app, or one MCP server. Codna sits where your engineering team already works.
One command.
codna fix . in any repo. Or codna triage . to understand it instantly.
codna triage . codna fix . --issue "checkout fails"
In your editor.
Add Codna to Cursor or Claude as an MCP server. Your agent gets a brain that actually knows the codebase.
Native GitHub App.
Install Codna on your repos, triage issues, open fix pull requests, and keep engineers in the review loop.
A patch you can trust.
Every fix comes with root cause, confidence, blast radius, and simulated regression risk.
Runs on your infrastructure.
Bring your own model key.
Fail-closed egress in private mode.
No training on your code. Ever.
Private by design
Your code is yours. Full stop.
Codna is built on a privacy-first engine: per-tenant key isolation, secret redaction, and audited egress. The guarantees are not a promise — they are enforced in the runtime.