Interview-driven, research-first API development toolkit with 14-phase TDD workflow, enforcement hooks, and 23 Agent Skills for cross-platform AI agents
npm install @hustle-together/api-dev-tools``
╔═══════════════════════════════════════════════════════════════╗
║ ║
║ ██╗ ██╗██╗ ██╗███████╗████████╗██╗ ███████╗ ║
║ ██║ ██║██║ ██║██╔════╝╚══██╔══╝██║ ██╔════╝ ║
║ ███████║██║ ██║███████╗ ██║ ██║ █████╗ ║
║ ██╔══██║██║ ██║╚════██║ ██║ ██║ ██╔══╝ ║
║ ██║ ██║╚██████╔╝███████║ ██║ ███████╗███████╗ ║
║ ╚═╝ ╚═╝ ╚═════╝ ╚══════╝ ╚═╝ ╚══════╝╚══════╝ ║
║ ║
║ API Development Tools for Claude Code ║
║ ║
╚═══════════════════════════════════════════════════════════════╝
Interview-driven, research-first API development with 14-phase TDD workflow
---
| Problem | Solution |
| --------------------------------------- | --------------------------------------------- |
| AI writes code from memory, not docs | ► Research-first workflow forces doc lookup |
| Generic questions miss project context | ► Interview questions generated FROM research |
| No verification after implementation | ► Phase 10 re-researches and compares to docs |
| Easy to skip TDD steps | ► 22 hooks enforce phase completion |
| Context dilutes over long conversations | ► Re-grounding every 7 turns |
| No visibility into AI workflow | ► State tracking + NTFY push notifications |
---
`bash`
npx @hustle-together/api-dev-tools --scope=project
The installer walks you through:
- ■ MCP server configuration (Context7, GitHub)
- ■ Environment variables (.env setup)
- ■ NTFY push notifications (optional)
- ■ Optional tools (Storybook, Playwright, Sandpack)
---
All four workflows share the same 14-phase structure ensuring consistency across API, component, page, and orchestration development.
Creates a complete API endpoint with research-backed implementation:
`bash`
/api-create stripe-checkout
Flow: Disambiguate → Research Stripe docs → Interview about error handling, formats → Generate Zod schemas → Write failing tests → Implement → Verify against docs → Refactor → Document
---
Creates UI components with Storybook integration and visual testing:
`bash`
/hustle-ui-create ChatWindow
Flow: AI suggests Basic/Complex type → User confirms → Research component patterns → Interview about variants, states → Generate types → Write Storybook stories → Implement → Visual regression tests → Document
---
Creates Next.js pages with Playwright E2E testing:
`bash`
/hustle-ui-create-page Dashboard
Flow: Same 14 phases but focused on page routing, data fetching, and E2E user flows instead of component isolation.
---
Combines multiple existing APIs into orchestrated endpoints:
`bash`
/hustle-combine api
Flow: Select from registry → Define orchestration pattern → Generate combined schemas → Test integration → Document the combined endpoint
---
`
┌─────────────────────────────────────────────────────────────────────┐
│ PHASE 1-2: CLARIFY │
│ ├─ Disambiguation: "Did you mean X or Y?" │
│ └─ Scope: "We're building Z with these features" │
├─────────────────────────────────────────────────────────────────────┤
│ PHASE 3-5: RESEARCH │
│ ├─ Initial: Context7 + WebSearch for official docs │
│ ├─ Interview: Questions FROM discovered parameters │
│ └─ Deep: Propose additional searches based on answers │
├─────────────────────────────────────────────────────────────────────┤
│ PHASE 6-7: PREPARE │
│ ├─ Schema: Zod types from research + interview decisions │
│ └─ Environment: Verify API keys exist before coding │
├─────────────────────────────────────────────────────────────────────┤
│ PHASE 8-9: BUILD (TDD) │
│ ├─ Red: Write failing tests that define expected behavior │
│ └─ Green: Minimal implementation to pass tests │
├─────────────────────────────────────────────────────────────────────┤
│ PHASE 10-12: VERIFY & REVIEW │
│ ├─ Verify: Re-research docs, compare to implementation │
│ ├─ Code Review: Greptile AI review (catches issues early) │
│ └─ Refactor: Fix issues + clean up while tests stay green │
├─────────────────────────────────────────────────────────────────────┤
│ PHASE 13-14: COMPLETE │
│ ├─ Documentation: Update registry, cache research, TypeDoc │
│ └─ Completion: Final verification, commit, PR │
└─────────────────────────────────────────────────────────────────────┘
---
| Component | Count | Purpose |
| ----------------- | ----- | --------------------------------------- |
| Agent Skills | 24 | Cross-platform workflow commands |
| Enforcement Hooks | 22 | Block progress until phases complete |
| Subagents | 7 | Parallel research, schemas, tests, docs |
| State Tracking | 1 | Track progress across sessions |
| Research Cache | 1 | 7-day freshness documentation cache |
| Templates | 4 | Component, page, env, MCP configs |
---
Seven specialized agents run in parallel to speed up workflows:
| Agent | Model | When Used |
| ----------------------- | ------ | ------------------------------------ |
| parallel-researcher | Haiku | Phase 3/5 - Scrape docs in parallel |
| research-validator | Haiku | Phase 3/5 - Find endpoints, webhooks |
| schema-generator | Sonnet | Phase 6 - Generate Zod schemas |
| test-writer | Sonnet | Phase 8 - Write comprehensive tests |
| implementation-reviewer | Sonnet | Phase 10 - Compare code to docs |
| code-reviewer | Sonnet | Phase 11 - Greptile AI code review |
| docs-generator | Haiku | Phase 13 - TypeDoc generation |
---
Get notified on your phone when input is needed:
`bash`
/ntfy-setup # Configure NTFY
/ntfy-test # Send test notification
Notifications include:
- ○ Input needed (interview questions)
- ● Phase completions with token usage
- ◆ Workflow blocked (missing requirements)
---
Interactive documentation and testing UI for all your APIs:
`bash`Access at: http://localhost:3000/api-showcase
Features:
- Grid view of all registered APIs with search and filtering
- Click any API to open interactive testing modal
- Example Requests - Pre-built, runnable examples that auto-fill query params
- Try It - Live API testing with real responses
- Curl Examples - Copy working curl commands
- Documentation - File locations, schemas, parameters
Examples are generated from your Zod schema parameters during Phase 13 (Documentation):
`json`
// In registry.json - auto-generated from schema
"endpoints": {
"search": {
"params": [...],
"examples": {
"basic": {
"description": "Basic search request",
"query": "action=search&query=nature",
"curl": "curl -X GET 'http://localhost:3000/api/v2/unsplash?action=search&query=nature'"
}
}
}
}
The showcase reads examples from registry.json and displays clickable buttons that auto-fill the request.
---
Generate API documentation from TSDoc comments:
`bash`
pnpm typedoc # Generate docs to docs/api/
pnpm typedoc:watch # Watch mode for development
TypeDoc runs during Phase 13 (Documentation) and generates Markdown documentation from:
- src/lib/schemas/*.ts - Zod schemas with TSDoc commentssrc/app/api/*/.ts
- - API route handlers
Configuration: typedoc.json (installed by the CLI)
---
`bash`
/red # Write ONE failing test
/green # Minimal implementation to pass
/refactor # Clean up while tests pass
/cycle [desc] # Complete Red → Green → Refactor
`bash`
/commit # Semantic commit with attribution
/pr # Create pull request
/busycommit # Multiple atomic commits
`bash`
/plan [feature] # Create implementation plan
/gap # Find unaddressed items
/issue [url] # Plan from GitHub issue
---
| Document | Purpose |
| -------- | ------- |
| docs/SKILLS.md | Complete slash command reference - all 24+ skills with usage and examples |
| docs/HOOKS.md | All 45+ enforcement hooks - what they do and when they run |
| docs/AGENTS.md | Specialized subagents - parallel-researcher, schema-generator, etc. |
| docs/ORCHESTRATOR.md | Master workflow controller - /hustle-build, decomposition, shared decisions |
| docs/REGROUNDING.md | 7-turn context refresh system - prevents "lost in the middle" problem |
| docs/PLUGIN_ARCHITECTURE.md | How the plugin system works - installation, state, lifecycle |
| docs/CLAUDE_CODE_BEST_PRACTICES.md | Industry best practices for Claude Code - hooks, subagents, context engineering |
| docs/GAP_ANALYSIS.md | How api-dev-tools implements best practices and what gaps remain |
| Document | Purpose |
| -------- | ------- |
| INSTALLATION_GUIDE.md | Step-by-step setup including MCP config, environment variables, troubleshooting |
| BEST_PRACTICES_ANALYSIS.md | Phase documentation with workflow diagrams and enforcement logic |
| .skills/README.md | Agent Skills format compatibility and cross-platform notes |
| Document | Purpose |
| -------- | ------- |
| CHANGELOG.md | Version history with features and release notes |
| ROADMAP.md | Future features and enhancement plans |
---
| File | Purpose |
| ---------------------------- | -------------------------------------------- |
| .claude/settings.json | Hook registration, tool permissions |.claude/api-dev-state.json
| | Current workflow state, phase progress |.claude/research/
| | Cached documentation with freshness tracking |.claude/registry.json
| | All created APIs, components, pages |templates/.env.example
| | Environment variable template |
---
- Claude Code 1.0.0+ (or compatible platform)
- Node.js 18+
- Python 3.9+ (for hooks)
- pnpm 10.11.0+
---
`bash``
npx @hustle-together/api-dev-tools --with-storybook # Component development
npx @hustle-together/api-dev-tools --with-playwright # E2E testing
npx @hustle-together/api-dev-tools --with-sandpack # Live code editing
---
- Issues: GitHub Issues
- Discussions: GitHub Discussions
---
License: MIT | Author: Hustle Together