Docs Overview

This page is the high-level guide to TestGen. It explains what the tool is, how to get started, the main ways to use it, and where to find the current source-of-truth docs in the repository.

TestGen is a multi-language tool for inspecting code, generating tests, validating coverage, and supporting both human and agent-driven workflows through the TUI, CLI, repo-local wrappers, and an MCP server.

Installation

Install TestGen locally, then set one provider key before generating tests.

curl -fsSL https://raw.githubusercontent.com/princepal9120/testgen/main/install.sh | bash
irm https://raw.githubusercontent.com/princepal9120/testgen/main/install.ps1 | iex
go install github.com/princepal9120/testgen-cli@latest
Compatibility note:
- GitHub repo: github.com/princepal9120/testgen
- Go module:  github.com/princepal9120/testgen-cli

Configuration

Set one provider API key before generating tests. TestGen supports Anthropic, OpenAI, Gemini, and Groq.

# Choose one provider
export ANTHROPIC_API_KEY="sk-ant-..."
export OPENAI_API_KEY="sk-..."
export GEMINI_API_KEY="AIza..."
export GROQ_API_KEY="gsk_..."
  • Anthropic Claude: strong default quality for many repositories
  • OpenAI: common general-purpose option
  • Gemini: useful when you want a Google-hosted option
  • Groq: useful when low latency matters

Quick Start

Keep the workflow simple: inspect first, preview generation second, write and validate last.

1. Inspect a codebase

testgen analyze --path=./examples --cost-estimate --recursive

2. Preview generated tests without writing

testgen generate --file=./examples/python/calculator.py \
  --type=unit \
  --dry-run \
  --emit-patch \
  --output-format json

3. Generate and validate when ready

testgen generate --path=./src --recursive --type=unit --validate

4. Prefer a guided UI?

testgen tui

Usage Modes

TestGen supports four practical usage modes today:

  • TUI: best for first-time human users who want a guided terminal flow
  • CLI: best for local development, scripting, and CI automation
  • Agent wrappers: best when coding agents should inspect dry-run artifacts before applying writes
  • MCP: best when your client prefers tool calls over shell execution

Review-first Workflow

The clearest default workflow is also the safest one: analyze the repo, generate dry-run output, inspect the results, and only then write files with validation. 

# 1) inspect the repo
testgen analyze --path=./src --cost-estimate

# 2) preview structured output
testgen generate --file=./src/utils.py --type=unit --dry-run --emit-patch --output-format json

# 3) write files and validate later
testgen generate --path=./src --recursive --type=unit --validate

Integrations

All current integrations rely on the same app layer and the same review-first model.

  • Codex: .codex/skills/testgen/SKILL.md
  • Claude Code: .claude/commands/testgen.md
  • OpenCode: .opencode/commands/testgen.md
  • MCP: testgen mcp exposes testgen_generate, testgen_analyze, and testgen_validate

Recommended safe mode for agents

testgen generate --file=./src/utils.py --type=unit --dry-run --emit-patch --output-format json

Install wrappers into another repo

./scripts/install-agent-integrations.sh /path/to/target-repo copy

Supported Languages

TestGen currently detects and generates tests for these languages:

  • JavaScript / TypeScript: Jest, Vitest, Mocha
  • Python: pytest, unittest
  • Go: testing, testify
  • Rust: cargo test
  • Java: JUnit 5-oriented generation via the Java adapter