A fast-first multi-agent AI coding harness with intelligent model routing, configurable testing, and token-saving optimizations for rapid SaaS MVP development.
npm install -g imhcodeNavigate to your project directory and run the initialization wizard:
cd your-project
imhcodeThis scans your system for local engines (Claude, OpenCode, Codex, Antigravity, QwenCode, MimoCode) and sets up the Intelligent Model Routing configurations inside imhcode.config.json.
IMH-Code enforces a fast-first software delivery pipeline:
- Write: Describe your product requirements in
docs/start.md. - Plan: Run
imhcode planto generatedocs/brainstorming.md. - Brainstorm: Review recommended answers in
docs/brainstorming.mdand runimhcode planagain. - Sprint Plan: Sprints, tasks, and shell execution scripts are generated.
- Execute: Run
imhcode execute 1(or specify sprint). Respects your selected testing mode. - Test: Run
imhcode testto execute E2E browser tests, security audit, and SEO scans.
Unlike other frameworks that mandate slow TDD at every step, IMH-Code lets you choose:
- Fast Mode (Default): No tests written during development sprints. A dedicated testing sprint is appended at the end to cover unit, integration, security, and browser tests. 3-5x faster development.
- Balanced Mode: Basic smoke tests per task, full test suites at the end.
- Strict TDD: Write failing tests first before coding on every task.
IMH-Code replaces individual persona names with generic agents routed to the best models:
- planner: Sprint coordination and architecture.
- designer: UI/UX design (prefer Mimo v2.5 Pro).
- nextjs-executor: Next.js full-stack (prefer Mimo v2.5 Pro).
- laravel-executor: Laravel backend (prefer DeepSeek V4 Pro).
- tester: Unit/Integration/E2E tests (prefer GPT-5.5).
- security-auditor: OWASP security scans (prefer GPT-5.5).
- seo-optimizer: SEO & sitemaps (prefer GPT-5.5).
- (Run
imhcode agent listto see all 19 agents)