tsforge wraps any OpenAI-compatible model in a strict gate: tsc, type-aware ESLint, 184 hand-tuned rules, and your own tests all have to pass before a change counts as done. No as any, no @ts-ignore, no escape hatches. Every patch is type-checked in-process before it touches disk, and the loop keeps going on its own until the build is green. Runs local or hosted, on large or small models. Open source, MIT.
No reviews yetBe the first to leave a review for tsforge - full stack typescript harness
Maker
📌
Hey Product Hunt 👋
I build TypeScript for a living, and I got tired of AI coding tools that produce code that looks right and quietly leaks any, @ts-ignore, and untested paths. The model moves fast; the cleanup lands on you.
tsforge flips that. It puts the model inside a gate. Before any change is "done," it has to pass tsc strict, type-aware ESLint, 184 hand-tuned rules, and your test suite. If something fails, tsforge re-plans and retries on its own, bounded by a turn budget. Green is the only exit.
A few things I care about:
- Zero escape hatches. as any, as unknown as, @ts-ignore, floating promises, and friends are banned, not warned.
- Reads your editor's truth. Patches are checked against an in-process TypeScript language service before they hit disk, so a patch that wouldn't compile never lands.
- Model-agnostic. Any OpenAI-compatible endpoint, local or hosted. It's tuned to work even with small local models, not just frontier ones.
- Stack-aware. Rule packs for React, Next, TanStack Query, Drizzle, Fastify, Elysia, BullMQ, and more.
- Composable. One CLI for chat, review, map, a merge gate, and MCP.
It's open source (MIT).
I'd genuinely love feedback on the rule set and where the gate is too strict or not strict enough.
Ask me anything.