zond

AI-powered API testing for Claude Code, Cursor, and CI/CD.

Say "test my API" — get working tests, coverage dashboard, and CI config in minutes.

Zond reads your OpenAPI spec and gives your AI agent everything it needs to test your API: a focused CLI, safety guardrails, coverage tracking, and run history. You don't need to learn anything new — just describe what you want and the agent runs zond commands.

Quick Start

Install the binary (no Node.js required):

curl -fsSL https://raw.githubusercontent.com/kirrosh/zond/master/install.sh | sh   # macOS/Linux
iwr https://raw.githubusercontent.com/kirrosh/zond/master/install.ps1 | iex        # Windows

Bootstrap a workspace and register your first API:

zond init --workspace --with-spec ./openapi.json

zond init writes a self-contained AGENTS.md — agents read it and use the CLI directly (zond run, zond probe-validation, zond db diagnose, …). No daemon, no transport, no extra configuration.

Then say to your agent: "Safely cover the API from openapi.json with tests."

npx -y @kirrosh/zond --version

See ZOND.md for the full CLI reference.

What Happens

1. Point — you give the agent an OpenAPI spec
2. Generate — zond reads the spec, produces YAML test suites (smoke + CRUD)
3. Run — tests execute, failures are diagnosed, coverage is tracked

The agent does all three steps autonomously. It asks you only when it needs an auth token or permission to run write operations.

Why Not Just Ask Claude to Write pytest?

Claude Code can write pytest from scratch — but it takes 30-60 minutes per flow, has no safety guardrails, no coverage tracking, and no run history. Zond gives the agent structured tools to do it in 5 minutes with full visibility.

Key Capabilities

| | | |---|---| | Safe by Default | --safe runs only GET requests. --dry-run previews without sending. The agent never touches production data without your explicit approval. | | Spec-Grounded | Tests are derived from your OpenAPI schema, not invented from scratch. The spec is the source of truth. | | Full Visibility | Every run is stored in SQLite. Compare runs, track regressions, see exactly what the server returned. | | Coverage Tracking | See which endpoints are tested, which aren't, and what broke since last run. | | CI-Ready | One command generates GitHub Actions or GitLab CI workflow. Tests in YAML, in git, with code review. |

Try It

"Cover openapi.json with tests"
"Run only smoke tests against staging"
"What broke since last run?"
"Set up CI for API tests"

Shell completions

zond completions bash > ~/.local/share/bash-completion/completions/zond
zond completions zsh  > ~/.zsh/completions/_zond   # then `compinit`
zond completions fish > ~/.config/fish/completions/zond.fish

Documentation

• ZOND.md — full CLI reference
• docs/quickstart.md — step-by-step quickstart (RU)
• docs/ci.md — CI/CD integration
• backlog/ — project tasks (powered by Backlog.md, see docs/backlog.md)

License

MIT