🥖 baguette

The Bun + Hono API framework your AI agent can't make a mess of.

Generated backend code rots because there are a hundred ways to do everything. baguette ships one — file-based routes, zod-first I/O, batteries included — and enforces it with a checker your AI can't merge around. Blazingly fast, zero-config, typed end to end.

Docs · npm · llms.txt

Quickstart

bun create baguette my-api
cd my-api && bun run dev     # → http://localhost:3000 · docs at /api/docs

Add an endpoint — the file's location is the URL, and one declaration gives you validation, inferred types, OpenAPI docs, and an error funnel:

// api/customers/[id].ts  →  GET /api/customers/{id}
import { defineRoute, z } from "@prehoy/baguette";

export default defineRoute({
  method: "get",
  request: { params: z.object({ id: z.string() }) },
  response: z.object({ id: z.string(), name: z.string() }),
  handler: (c, { params }) => c.json({ id: params.id, name: "Ada" }),
});

// server.ts
import { serve } from "@prehoy/baguette";
serve({ routesDir: "./api" });   // routes loaded, docs live, done

Built for AI agents

The thing that makes generated code a mess — infinite ways to structure it — is the thing baguette removes. Every app on baguette is the same shape, so an agent (or a new hire) can't drift:

• One obvious way — one route per file, all I/O through zod, no manual parsing.
• A shipped clean-code contract every repo inherits — read by humans and the AI tools working in it.
• baguette/eslint + baguette check turn the conventions into CI failures, so an agent physically can't merge the mess.
• llms.txt — the whole framework, machine-readable, so agents get it right the first time.

Batteries included — each opt-in

| | | |---|---| | Routing | File-based, [id] params, method-suffixed files, auto Scalar docs | | Validation | zod-first: one schema → validation + types + OpenAPI | | Security | Declarative per-route auth, rate limiting, security headers, CORS guard, body limits | | Real-time | Opt-in WebSockets with a room/channel pub/sub | | Background | cron/ (memory/Postgres/Redis/SQLite locks), queues/ (bee-queue), automations/ (LISTEN/NOTIFY) | | Email | React templates, browser preview, send via Resend/SMTP | | Ops | onBoot/onShutdown + graceful SIGTERM, static/SPA serving, typed defineEnv | | Agnostic | No ORM in the core — Prisma, Drizzle, or raw SQL |

Full docs: usebaguette.com/docs

Deploy

bun create baguette ships a Dockerfile — run it anywhere, or have it managed on Berth. Building something bigger? Talk to Prehoy — we do the infra and the backend.

Contributing / dev

bun install
bun test
bun run dev      # runs examples/ on :3000

MIT · a Prehoy Industries project · usebaguette.com