Package Exports

This package does not declare an exports field, so the exports above have been automatically detected and optimized by JSPM instead. If any package subpath is missing, it is recommended to post an issue to the original package (tlc-claude-code) to support the "exports" field. If that is not possible, create a JSPM override to customize the exports field for this package.

Readme

TLC

Test Led Coding. Tests before code. Automatically.

npx tlc-claude-code

TLC

The Problem

You tell Claude to build something. It builds it. You test it manually. It's broken. You debug. Repeat.

That's backwards.

The Solution

TLC writes tests before code exists. Every feature has a spec. Every spec is executable. When the code works, you know — because the tests pass.

You describe → Tests are written → Code is implemented → Tests pass → Done

No manual testing. No "does this work?" No vibes.

Quick Start

npx tlc-claude-code        # Install

Then in Claude Code:

/tlc

That's it. One command. It knows what to do next.

Starting fresh? It asks what you're building. Have existing code? It finds untested files. Mid-project? It picks up where you left off.

Commands

Command	What
`/tlc`	Smart entry point. Knows what's next.
`/tlc:new-project`	Start fresh. Discuss stack, scaffold.
`/tlc:init`	Add TLC to existing code.
`/tlc:coverage`	Find untested → write tests
`/tlc:quick`	One-off task with tests
`/tlc:status`	Pass/fail counts

What Makes This Different

1. Tests First, Always

Other workflows: plan → build → "hope it works"

TLC: plan → write failing tests → build until tests pass

The tests are the spec. No ambiguity.

2. Smart Stack Selection

Don't pick tech in a vacuum. TLC asks what you're building, who uses it, what scale — then suggests the right stack.

Building: Internal dashboard
Scale: Small team
Data: Simple CRUD

→ Suggested: Next.js + SQLite + Vercel
→ Why: Fast to build, cheap to host, fits your needs

3. Parallel Agents

Up to 3 Claude instances working simultaneously. GitHub issues as task queue. Watch them go.

┌──────────────────────────────────────────────────────┐
│ Agents                                               │
│ [1] ● Working on #42: Auth flow                      │
│ [2] ● Working on #43: User CRUD                      │
│ [3] ○ Idle                                           │
└──────────────────────────────────────────────────────┘

4. GitHub Integration

Plans approved → issues created automatically. Tasks complete → issues closed. Full audit trail.

5. Live Preview

Docker container spins up. See your app as it's built. Not after.

Dashboard (Coming Soon)

┌─────────────────────────────────┬──────────────────────┐
│ Chat                            │ GitHub Issues        │
│                                 │ #42 Auth flow    WIP │
│ Building login endpoint...      │ #43 User CRUD        │
│ ✓ Created tests/auth.test.ts    │ #44 Dashboard        │
│ ✓ Tests failing (expected)      ├──────────────────────┤
│ Implementing...                 │ Agents (2/3)         │
│                                 │ [1] ● #42            │
│                                 │ [2] ● #43            │
│                                 │ [3] ○ Idle           │
├─────────────────────────────────┼──────────────────────┤
│ > add password reset flow       │ Tests: 23/23 ✓       │
└─────────────────────────────────┴──────────────────────┘

TUI dashboard. Multiple panes. Real-time updates.

Philosophy

Tests define behavior. Code makes tests pass.

Tests written BEFORE code
Tests are the spec, not an afterthought
If it's not tested, it doesn't exist
Human verification still happens — tests catch logic errors, you catch "not what I meant"

vs Other Approaches

Approach	Process	Result
Vibe coding	Build → hope	Works until it doesn't
Manual TDD	Write tests yourself	Slow, easy to skip
TLC	Tests auto-generated first	Fast, guaranteed coverage

Install

npx tlc-claude-code

Options:

--global — Available everywhere
--local — This project only

License

MIT