# Eval-Driven Development

> TDD for AI agents that call tools. Write evals, mock MCP schemas, assert routing, gate PRs, and turn production misses into the next case.

## What this site is
Eval-Driven Development (EDD) is the same red → green → refactor loop as TDD, applied to prompts and tool schemas. The harness ships in Agent Lifecycle Kit (`kit eval run|ci|report`). The same kit also gates always-on context (`kit measure-context`), composes one MCP profile, and runs `kit check`.

## What do I use this for today?
- Wrong tool / made-up args → `kit eval run --suite evals/edd/demo.yaml --model scripted`
- Gate a prompt or schema change → `kit eval ci --suite evals/edd/demo.yaml --threshold-routing 95 --out out/reports`
- Always-on context too fat → `kit measure-context` then `kit check`
- Starting a product feature → lifecycle (grill → spec → TDD + XFN → audit → release)
- Never installed → start here in 10 minutes: install → `kit init` → demo suite → CI bar

## Start here
- [Home](https://eval-driven.dev/)
- [Getting started](https://eval-driven.dev/docs/start) ([markdown](https://eval-driven.dev/docs/start.md))
- [Jobs for today](https://eval-driven.dev/docs/jobs) ([markdown](https://eval-driven.dev/docs/jobs.md))
- [Common questions](https://eval-driven.dev/docs/faq) ([markdown](https://eval-driven.dev/docs/faq.md))
- [Docs overview](https://eval-driven.dev/docs) ([markdown](https://eval-driven.dev/docs/README.md))
- [EDD guide](https://eval-driven.dev/docs/edd) ([markdown](https://eval-driven.dev/docs/edd.md))
- [Operator guide](https://eval-driven.dev/docs/kit) ([markdown](https://eval-driven.dev/docs/kit.md))
- [Feature lifecycle](https://eval-driven.dev/docs/lifecycle) ([markdown](https://eval-driven.dev/docs/lifecycle.md))
- [Kit map](https://eval-driven.dev/docs/map) ([markdown](https://eval-driven.dev/docs/map.md))
- [SOPs](https://eval-driven.dev/docs/sops) ([markdown](https://eval-driven.dev/docs/sops.md))
- [Demo suite](https://eval-driven.dev/evals/edd/demo.yaml)
- [Before/after write-up](https://eval-driven.dev/evals/edd/examples/before-after) ([markdown](https://eval-driven.dev/evals/edd/examples/before-after.md))
- [Eval suites](https://eval-driven.dev/evals/edd) ([markdown](https://eval-driven.dev/evals/edd/README.md))
- [GitHub](https://github.com/mzworthington/agent-lifecycle-kit)

## Install
```
curl -fsSL https://raw.githubusercontent.com/mzworthington/agent-lifecycle-kit/main/install.sh | sh
kit init . --mcp default --hook
```

## Commands
```
kit eval run --suite evals/edd/demo.yaml --model scripted
kit eval ci --suite evals/edd/demo.yaml --threshold-routing 95 --out out/reports
kit eval report --format md --out out/reports
kit measure-context
kit check
kit mcp default --install
```