frogbot.config.ts — FrogBot boots the production agent backend for all of it. No routing code, no glue, no SaaS.
That one file gives you a hosted UI, agent runtime, chat, skills, flows, integrations, model routing, file storage, authentication, REST and GraphQL APIs, and a job system — all in one MIT open-source TypeScript codebase you own and deploy anywhere.
Start here
Paste this prompt into Codex, Claude Code, Cursor, or another coding agent, and it sets up FrogBot for you:Your coding agent checks your computer, creates the app with the FrogBot skill, and starts it.
Step-by-step setup
Go from nothing installed to your first chat, one plain-language step at a time.
Start with a coding agent
What the prompt asks your agent to do, and what it will ask you.
Installation
Create a project with
npx create-frogbot-app@latest and choose its database and AI provider.Concepts
The high-level building blocks: config, models, pieces, skills, agents, flows, plugins, and content directories.
Configuration
The
frogbot.config.ts file that controls every part of the framework.Deployment
Run your app in production with Docker or Fly.io.
What it gives you
- An AI workspace UI for chats, agents, skills, flows, and connections
- A file-based defaults system for skills, agents, MCPs, and flows
- A pieces system for integrations (Slack, Gmail, Notion, GitHub, and more)
- A piece connector layer that handles managed OAuth without per-piece OAuth apps
- A job system for triggers, polling, scheduled flows, and waitpoints
- Web, desktop, and browser extension clients sharing the same React app shell
- A plugin system that extends the framework through config transforms