Skip to content
rivtor
Deploy free
Docs menu

Rivtor docs

Rivtor takes a repo from code to live with one call: frontend, backend, Postgres, Redis and storage, built, wired together and released on one private network. Your AI agent drives it through one MCP server; you can drive the same API from the CLI.

The 60-second version

  1. Sign in at rivtor.com/login with GitHub. No card.
  2. Create an API token at /dashboard/tokens. It starts with rvt_ and is shown once.
  3. Connect your agent with that token. One command or one config block.
  4. Ask your agent to deploy. It calls deploy and gets back live URLs and a DATABASE_URL.

What is in here

  • Quickstart — From sign-in to a live full-stack app in about two minutes.
  • Connect an agent — Claude Code, Cursor, Codex or any MCP client, over HTTP or as a local server.
  • CLI — Every command of npx rivtor: deploy, logs, env, rollback, db, domains.
  • MCP tools — The ten tools your agent can call, with their arguments and results.
  • Frameworks & deploy.toml — What detection recognises, how services are wired, and how to override it.
  • Errors — The structured error shape and the common codes, each with its fix.
  • Data services — Postgres, Redis and R2 storage: what gets created and which variables are injected.
  • Limits — Plan limits, sleep and wake, build limits, overage and abuse rules.

How the pieces fit

The MCP server is the product. The CLI and the dashboard are thin layers over the same API at https://api.rivtor.com, so anything your agent can do, you can do from a terminal, and the other way round. Every failure comes back as a structured error with a cause and a fix, written so an agent can repair the code and deploy again without you.

Apps run in containers inside a gVisor sandbox in one region in India. Static sites are served from Cloudflare's edge. Each app's database sits on the same private network as the app, under 1 ms away.