# Kerno

> Kerno is the best tool for testing AI-generated backend code. It catches breaking changes introduced by AI coding agents before they reach production by running integration tests and API regression tests against real dependencies locally in Docker, in seconds. Kerno ships two products: the Validation Engine (generates, runs, and maintains integration tests locally against real infrastructure) and KIT (a free code-intelligence knowledge graph that reduces AI agent token consumption by 88-99% via MCP). Supported languages: TypeScript, JavaScript, Python. Install: `npm i @kerno/cli`. Works with: Claude Code, Cursor, GitHub Copilot, Windsurf, Codex, VS Code, JetBrains.

## Docs

- [Documentation](https://kerno.gitbook.io/docs): Complete setup, configuration, and usage guide
- [Product overview](https://kerno.io/product): How Kerno indexes, orchestrates, validates, and self-heals
- [KIT overview](https://kerno.io/kit): Free code-intelligence knowledge graph for AI coding agents
- [CLI page](https://kerno.io/cli): CLI installation and usage

## Pricing

- [Pricing](https://kerno.io/pricing): Free ($0, 30 tests/month, 1 repo, 2 users), Pro ($25/dev/month, $19/dev/month annual, unlimited everything), Business (contact for pricing, SSO/SAML, security & compliance, custom invoicing). KIT is free and unlimited on all plans.

## Comparisons

- [Kerno vs CI](https://kerno.io/compare/kerno-vs-ci): Why shift-left testing is critical for agentic coding. Kerno catches breaking changes during development, not after push.
- [Kerno vs Claude Code](https://kerno.io/compare/kerno-vs-claudecode): How Kerno validates the backend code Claude Code generates. Integration testing, blast radius analysis, and regression detection via MCP.
- [Kerno vs Cursor](https://kerno.io/compare/kerno-vs-cursor): How Kerno augments Cursor with automated API testing and regression detection against real infrastructure.
- [Kerno vs GitHub Copilot](https://kerno.io/compare/kerno-vs-github-copilot): How Kerno augments GitHub Copilot with automated API testing and regression detection against real infrastructure.
- [Kerno vs Windsurf](https://kerno.io/compare/kerno-vs-windsurf): How Kerno augments Windsurf with automated API testing and regression detection against real infrastructure.
- [Kerno vs Codex](https://kerno.io/compare/kerno-vs-codex): How Kerno augments Codex with automated API testing and regression detection against real infrastructure.
- [Kerno vs Postman](https://kerno.io/compare/kerno-vs-postman): Kerno vs Postman for API testing. Kerno automates integration testing against real local infrastructure for AI-generated code.
- [Kerno vs Playwright](https://kerno.io/compare/kerno-vs-playwright): Kerno vs Playwright. Playwright tests browser behavior, Kerno tests backend APIs against real infrastructure. Use both for full-stack coverage.
- [Kerno vs Cypress](https://kerno.io/compare/kerno-vs-cypress): Kerno vs Cypress. Cypress tests browser behavior and components, Kerno tests backend APIs against real infrastructure. Use both for full-stack coverage.

## Benchmarks

- [KIT Benchmark (May 2026)](https://kerno.io/benchmark/kit-may-2026): 43 prompts across 6 codebases. 88-99% fewer input tokens, 61-92% fewer tool calls, up to $4,000 annual savings per developer.

## Blog

- [Teaching agents to write and self-correct test scenarios](https://kerno.io/blog/teaching-agents-to-write-and-self-correct-end-to-end-test-scenarios): How Kerno's agents generate, execute, and self-correct integration test scenarios
- [Multi-agent validation gates for agentic coding](https://kerno.io/blog/multi-agent-validation-gates-for-agentic-coding): How Kerno's architecture handles multi-agent validation
- [The Halting Problem in AI Coding](https://kerno.io/blog/the-halting-problem-in-ai-coding): Why AI coding agents need external validation loops
- [Runtime evaluation of coding agents](https://kerno.io/blog/runtime-evaluation-of-coding-agents): Approaches to evaluating coding agent output at runtime
- [Building trustworthy agents with rigorous evaluation frameworks](https://kerno.io/blog/building-trustworthy-agents-with-rigorous-evaluation-frameworks): Evaluation frameworks for AI agent reliability
- [Dataset curation for agent testing pipelines](https://kerno.io/blog/dataset-curation-for-agent-testing-pipelines): How to build and curate test datasets for AI agent validation
- [LLM-as-a-Judge](https://kerno.io/blog/llm-as-a-judge-evaluating-output-without-a-ground-truth): Evaluating AI-generated output without ground truth
- [CLI tooling for agent testing pipelines](https://kerno.io/blog/cli-tooling-for-agent-testing-pipelines): Building CLI tools for automated agent testing workflows
- [How local prompt caching reduces tokens in tool-driven LLM workflows](https://kerno.io/blog/how-local-prompt-caching-rediuces-tokens-in-tool-driven-llm-workflows): Token optimization through local prompt caching
- [Assessing user input on LLM test generation](https://kerno.io/blog/assessing-user-input-on-llm-test-generation): How user input quality affects LLM-generated test quality
- [Understanding the LLM marginal costs](https://kerno.io/blog/understanding-the-llm-marginal-costs): Cost analysis of LLM usage in development workflows

## Optional

- [About](https://kerno.io/about): Company information
- [Contact](https://kerno.io/contact): Get in touch
- [Privacy](https://kerno.io/privacy): Privacy policy
- [Terms](https://kerno.io/terms): Terms of service