Two ways to run agents. Pick the mode that fits your workflow — or use both.
CLI Mode
Run agents locally with full file access. One command wraps any process in a streamable session — monitor from your phone, approve tool calls from the couch, keep agents productive while you do other things.
- +Full local file system access
- +Stream any terminal session to the dashboard
- +Multi-machine support — all in one view
- +PTY mode for shells, structured mode for AI agents
Server-Side SDK
Rich agent conversations, right from the browser
Server Mode runs Claude through a server-side SDK integration with full markdown, tool call blocks, and multi-turn conversations — no respawning between messages. CLI sessions render the same way, streamed from your local machine.
1.0 Server-Side SDK →Multi-Machine
All your machines, one dashboard
Sessions are automatically grouped by hostname. Run agents on your desktop, a remote server, and a cloud VM — monitor them all from a single view. CLI version tracking tells you when machines need updates.
2.0 Multi-Machine →Permissions
Stay in control without being in the terminal
Approve or deny tool calls per-session or per-project. Toggle between normal and YOLO mode on the fly. Grant blanket permissions for trusted tools, or review each action individually.
3.0 Permissions →Coming Soon
Multi-provider support
Server Mode will expand beyond Claude to support Gemini, Codex, and more. Run different providers side by side and pick the best tool for each task — all from the same dashboard.
Get Started
Choose your mode. Start in seconds.
Server Mode — no install needed
Agents run in a sandboxed server environment. Great for testing and experimentation.
CLI Mode — full local access
Agents run locally with full file access. Sessions stream to the dashboard automatically.
Built for developers who run AI agents — whether from the browser or the command line — and want to stay in control.
Quick experiments from the browser
Open the dashboard, start a Server Mode session, and test ideas in a sandboxed environment. No setup, no install — just start prompting.
Kick off agents before bed
Start a long refactor with CLI Mode from your laptop, check progress from your phone in the morning. Approve the one permission it's waiting on, and it continues.
Parallel agents, one view
Run server-side and local agents simultaneously. See all sessions in one dashboard. Know instantly which one needs your attention.
Start orchestrating your agents
Try Server Mode from the dashboard — no install needed. Or set up the CLI for full local access.
