Introduction
What agentstead.dev is and how the platform works in plain terms.
AgentStead gives you a full Linux desktop for your AI agent — persistent, always-on, and ready to use from your browser. No Docker stacks, no domain wiring, no infra to babysit.
Pick a runtime, hit deploy, and you get:
- A live remote desktop you can open in your browser
- Persistent workspace storage that survives restarts
- Pre-installed AI coding tools: Claude Code, Codex, OpenCode, VS Code, Neovim
- Credentials, secrets, logs, and backups all in one dashboard
- Optional private networking via Tailscale
Each agent gets its own isolated desktop environment — not a shared VM, not a container you have to nurse. It's closer to a persistent remote workstation.
Platform overview
Who this is for
AgentStead is a good fit if you want an always-on agent environment but don't want to run the control plane yourself. It works well for:
- AI coding workflows (Claude Code, Codex, OpenCode)
- Browser-assisted research and automation
- Multi-agent setups where each agent needs its own isolated workspace
- Long-running background tasks that need a persistent environment
- API-oriented agent deployments (OpenFang with REST surface)
- Media processing pipelines (OpenFang with ffmpeg)
The four runtimes
AgentStead offers four runtime types, each with a different agent daemon alongside the same desktop and toolchain:
| Runtime | Best for | Token type |
|---|---|---|
| OpenClaw | General AI coding, Claude Code, Codex | Gateway Token |
| OpenFang | REST API + media workflows, no key required | API Key |
| NullClaw | Zig-based workloads | Session Token |
| ZeroClaw | Rust-based workloads | Session Token |
Not sure which to pick? Start with OpenClaw — see Choose Your Runtime.