ProBot

Self-hosting guide

Run your bot on your own infra.

Self-hosting in ProBot means running your bot's chat on your own infrastructure - not operating the whole platform. You deploy the tiny probot-bot runtime under your own domain; pro-bot.dev keeps handling the dashboard, knowledge, conversations, and leads. Your LLM key lives in your runtime and never touches pro-bot.dev.

How it works

  1. A visitor chats with your probot-bot runtime on your domain.
  2. The runtime asks the platform for the relevant knowledge and your bot's persona over the versioned /api/v1/bot/* API, authenticated with a bot token.
  3. The runtime calls your LLM provider directly - using the key in your own environment - and replies.
  4. It posts the transcript and any captured lead back to the platform, so they appear in your ProBot dashboard exactly like a managed bot.

What you'll need

  • A ProBot account with a bot already created (build it in the dashboard first).
  • A Node 20+ host for the runtime - Vercel, Render, Fly.io, Railway, a VM, or Docker. The runtime is tiny, so a free tier is plenty.
  • An LLM API key from Anthropic, OpenAI, Google, or Azure.

Steps

  1. In the dashboard, open Settings → Deployment and switch the bot to Self-hosted.
  2. Generate a bot token. It's shown once - copy it somewhere safe.
  3. Deploy the probot-bot runtime with that token in its PROBOT_BOT_TOKEN environment variable and your LLM key in its provider key variable.
  4. Point your domain at the runtime. Revoke the token any time to instantly cut it off.

Why self-host

  • You want the chat served from your own domain.
  • You want zero trust in any operator for the chat path.
  • You want a tiny, auditable deployment surface you control.

Full setup, environment variables, and the API contract live in the self-hosting quickstart.

Questions about the deployment? See About for contact info.