Your AI coding assistant,
held to production standards.

ForgeCraft gives every AI assistant in your team a shared quality contract — architecture, testing, CI/CD, dev hygiene — from the first commit.

Get started → View on GitHub
$npx forgecraft-mcp setup .
You hired an AI engineer. It's brilliant. It also installed the same 14 VS Code extensions twice today, spun up 6 Docker containers it will never clean up, and your disk went from 12 GB free to 0 KB in one session.

A full disk doesn't fail gracefully. It kills VS Code, the terminal, Docker, and the database simultaneously.

What ForgeCraft does

It runs once at project setup and generates instruction files for your AI assistant — tailored to your stack — covering architecture, testing, security, deployment, and dev environment hygiene. The AI works within that contract for the life of the project.

116 curated instruction blocks

SOLID, hexagonal architecture, testing pyramid, CI/CD, clean code, domain patterns — assembled to match your project's tags.

7-property quality score

Every project scores 0–14 on the Generative Specification model. Not a linter score — a measure of whether the codebase can sustain AI-assisted development.

Quality gates by release phase

From unit tests on day one to OWASP and chaos engineering before release. Gates the library is community-contributed and open.

Dev environment hygiene

Rules injected into every project: no duplicate extensions, no orphaned containers, disk warnings before the workspace silently fills up.

ADRs, automatically sequenced

Every architectural decision recorded in MADR format. Your AI assistant reasons about past choices. Your team stops re-litigating them.

Session continuity

Status.md and forgecraft.yaml keep context across sessions and engineers. The AI always knows where the project is.

Quality score — out of 14

Run npx forgecraft-mcp verify . at any point to see where the project stands:

| Property | Score | Evidence | |-----------------|--------|-------------------------------------------------| | Self-Describing | ✅ 2/2 | CLAUDE.md — 352 non-empty lines | | Bounded | ✅ 2/2 | No direct DB calls in route files | | Verifiable | ✅ 2/2 | 64 test files — 87% coverage | | Defended | ✅ 2/2 | Pre-commit hook + lint config present | | Auditable | ✅ 2/2 | 11 ADRs in docs/adrs/ + Status.md | | Composable | ✅ 2/2 | Service layer + repository layer detected | | Executable | ✅ 2/2 | Tests passed + CI pipeline configured | Total: 14/14 ✅ PASS · Threshold 11/14

Works with every AI assistant

One setup. Instruction files generated for whichever assistant your team uses.

Claude (CLAUDE.md) GitHub Copilot (.github/copilot-instructions.md) Cursor (.cursor/rules/) Windsurf (.windsurfrules) Cline (.clinerules) Aider (CONVENTIONS.md)

Running this with a team? The Forge workshop puts this methodology into practice on your actual codebase — in two days.

See The Forge Workshop →

Research foundation

ForgeCraft implements the Generative Specification model — a 7-property framework for evaluating AI-generated code quality. Published, DOI-backed, independently verifiable. Read the white paper →

Open source. Free to use.

No limits, no tiers, no API keys. The quality gate library grows through community contribution.

Read the docs → Contribute a gate →