Skip to content
SAM Docs

What is SAM?

Simple Agent Manager (SAM) is an open-source platform for getting real work done with AI coding agents. You import a GitHub repository, describe what you want in a chat, and SAM runs your choice of AI coding agent — Claude Code, OpenAI Codex, Gemini CLI, Mistral Vibe, OpenCode, or Amp — that reads your code, makes the change, and opens a pull request. SAM provisions the environment for you.

Why SAM?

Section titled “Why SAM?”

AI coding agents need real development environments. SAM gives each agent an isolated environment behind a simple chat — a container that starts in seconds for conversation and focused edits, or a full cloud VM with your .devcontainer, Docker, and toolchain when the agent has to build and test your stack. You describe the outcome you want; SAM handles the servers, the setup, and the pull request.

Bring Your Own Cloud

Connect a Hetzner, Scaleway, Vultr, Infomaniak, DigitalOcean, UpCloud, or GCP credential and VMs run on your own infrastructure, billed to your account. On the hosted platform, compute can be provided for you. Your cloud credentials are never stored as environment variables.

Serverless Control Plane

The API runs on Cloudflare Workers — zero infrastructure to manage. D1 for storage, KV for sessions, R2 for binaries, Durable Objects for real-time chat.

Ephemeral by Design

Container sessions start in seconds and VM workspaces in minutes, and both tear down when you’re done. Warm node pooling reuses idle VMs for fast follow-up work.

Self-Hostable

Fork the repo, configure a GitHub Environment, then run Deploy Production. Pulumi + GitHub Actions provision the stack.

How It Works

Section titled “How It Works”
  1. Sign in with GitHub via the web UI
  2. Connect an AI agent in Settings → Connections (Claude Code, Codex, Gemini CLI, and more)
  3. Install the GitHub App on the repositories you want to use
  4. Import a project — link a GitHub repo from the Dashboard
  5. Chat — describe what you want; SAM provisions a workspace and runs the agent for you
  6. Review the PR — the agent commits, pushes, and opens a pull request when it’s done

Key Features

Section titled “Key Features”
  • Six AI agents — Claude Code, OpenAI Codex, Gemini CLI, Mistral Vibe, OpenCode, Amp
  • Chat-first UX — describe what you want in plain language and watch the agent work in real time
  • Agent profiles — save reusable agent + model + settings combinations and pick one per chat
  • Skills — layer reusable instructions and tools on top of a profile for specific kinds of work
  • Triggers — start agent work automatically from a schedule, a GitHub event, or an authenticated webhook
  • File browsing in chat — browse, view, and diff files without leaving the conversation
  • File upload & download — attach files to conversations and download from workspaces
  • Conversation forking — branch from any point to explore alternatives
  • Voice input & TTS — speak your prompts, listen to responses with a persistent audio player
  • Command palette — quick navigation via Cmd+K / Ctrl+K
  • In-app notifications — progress updates, agent requests, PR creation alerts
  • App deployments — create deployment environments, gate agent deploy access, and publish Docker Compose releases
  • Bring your own cloud — run on your own Hetzner, Scaleway, Vultr, Infomaniak, DigitalOcean, UpCloud, or GCP account, or use hosted compute
Quickstart Import a repo and get your first pull request in minutes.
AI Agents Choose an agent, connect credentials, and configure profiles.
Architecture Overview For self-hosters and the curious: how every system fits together.