for teams adopting Claude Code ยท Cursor ยท Copilot ยท Codex

Stop your coding agents from
repeating the same mistakes

agent-tuner logs your AI agent's failed code generations, distills them into project-specific rules and regression evals, and auto-updates your agent config โ€” a self-improving loop tied to your codebase, not a static prompt library.

# one install, lives next to your code npx agent-tuner init npx agent-tuner log --summary "agent used fetch() in a server action; should use the db client" npx agent-tuner distill # โ†’ learns a rule + a regression eval npx agent-tuner apply # โ†’ updates CLAUDE.md / AGENTS.md / .cursorrules npx agent-tuner eval run # โ†’ gates CI; non-zero if the mistake comes back

The feedback loop

Four commands. The rules your team learns the hard way become permanent.

Log

Capture a failed generation in one line โ€” manually, from a shell hook, or in CI.

Distill

Your LLM turns failures into a small set of sharp, durable rules + an eval each.

Apply

Rules written into a managed block in your agent config. Idempotent; never clobbers your prose.

Eval

Re-runs the failing prompt with the new rules. Exits non-zero on regression โ€” gate your pipeline.

Why agent-tuner

Built for the way real teams use coding agents.

๐Ÿงญ

Tied to your codebase

Rules come from real failures in your repo โ€” not a generic best-practices list that ignores your conventions.

โœ…

Regression evals

Every rule ships with a test that proves the mistake is fixed and stays fixed. CI-gating, so quality can't silently rot.

๐Ÿ”Œ

Works with every agent

CLAUDE.md, AGENTS.md, .cursorrules, copilot-instructions, Windsurf. Any LLM via your own gateway โ€” models stay swappable.

๐Ÿชถ

Zero dependencies

One `npx` command, no node_modules, tiny supply-chain surface. Installs in seconds into any repo.

๐Ÿ”

Git-native

Rules and evals are readable JSON you commit. Your team's hard-won knowledge becomes a versioned asset.

๐Ÿข

Share across the org

Sync rules between repos and roll up an org-wide ruleset so every team's agents get smarter together.

Pricing

Free forever for a single repo. Pay when you want your rules to travel.

Free
$0
โ‚น0 โ€” forever
  • 1 repo
  • Unlimited local rules & evals
  • All CLI features
  • CI eval-gating
  • Every agent + every LLM gateway
Get started
Team
$19 /mo per repo
โ‰ˆ โ‚น1,599 /mo per repo
  • Everything in Free
  • Sync rules across repos
  • Shared eval history
  • Team rule review
  • Priority support
Start Team
Org
$49 /mo
โ‰ˆ โ‚น3,999 /mo
  • Everything in Team
  • Org-wide shared ruleset
  • Cross-team rollups
  • Approval workflow
  • GST invoice (India)
Contact

Pricing shown in USD and INR. Checkout via Stripe (global) and Razorpay/UPI (India) โ€” launching shortly. Until then, the Free tier is fully usable today: npx agent-tuner init.