Skip to main content

​
What is Pylon?

Pylon is an agent-native full-stack framework: point a coding agent at it and it ships a real app — typed schema, row-level policies, server functions, live queries, auth, and native React SSR, all from a single Rust server — it runs your TypeScript and SSR on Bun — that you run and deploy with one command each. You (or your agent) write the schema, policies, and functions in TypeScript; Pylon gives you a typed client and runs the whole backend, with realtime sync built in — apps that need shared, synchronized state (dashboards, CRMs, collaborative tools, chat) work without a separate realtime layer. It’s production infrastructure, not a sandbox — SQLite or Postgres, real auth, one-command deploy — so what you build is something you’d run a business on, not a throwaway capsule. No glue between your database, your API, and your WebSocket server. Everything is one system.

Quickstart

Build a live app in five minutes.

Core concepts

Entities, policies, functions, live queries.

SSR

File-based routes, instant nav, image optimizer — Pylon replaces Next.js for full-stack apps.

Deploy

Self-host, Pylon Cloud, or Vercel + Pylon Cloud.

​
What you get

  • Typed schema — declare entities in TypeScript, get types everywhere.
  • Row-level policies — access rules next to your data, not in middleware.
  • Server functions — queries, mutations, actions, all RPC-callable from the client.
  • Live queries — db.useQuery(...) subscribes to results and updates on write.
  • Auth — email magic links, OAuth, guest sessions. No third-party SDK.
  • SSR — file-based React routes with <Link>, <Image>, and built-in Tailwind — no Next.js required.
  • One server — a Rust binary that runs your TypeScript and SSR on Bun, on one port. Deploy anywhere.

​
Why another backend framework?

Most realtime backends are three systems glued together: a database, an API server, and a realtime pub/sub layer. Keeping them in sync is a full-time job. Pylon collapses the stack. Writes go through the same runtime that serves reads and pushes updates. Live queries re-execute on the server when dependent rows change and stream diffs to subscribers. There’s no denormalization, no cache invalidation, no manual fan-out.

​
What it’s good for

  • SaaS apps with live dashboards
  • Internal tools, CRMs, ERPs
  • Collaborative editors
  • Multiplayer games and worlds
  • Anything where two browsers need to see the same thing, now

​
Storage

Pylon runs on SQLite or Postgres. SQLite is the default — a single file, zero setup, great for local dev and many production workloads (single-node, up to tens of GB). Switch to Postgres for horizontal scale, existing Postgres infra, or shared-nothing deployments. The schema, policies, functions, and client APIs are identical across both — set DATABASE_URL=postgres://... and Pylon uses the Postgres adapter.

​
How to read these docs

Start with the Quickstart to feel the shape of a Pylon app in five minutes. Then Core concepts walks through entities, policies, functions, and live queries in depth. SSR covers file-based routing, <Link>, <Image>, and the rest of the full-stack story. Deploy covers self-hosting, and Deploying to Vercel is the canonical Next.js + Pylon Cloud production stack. Browse examples on GitHub for end-to-end apps (CRM, ERP, chat, 3D world, SSR demos, ten+ total) you can clone.