Local-first agent runtime
Run agents where your work lives.
Guides for the Local Operator agent runtime: the terminal UI, the desktop app, workflow automations, remote access, and building on top of it.


Start here
What is Local Operator?Local Operator is an agent runtime that runs on your machine — one runtime every surface, from the terminal to your phone.Install & first runInstall the Local Operator runtime with pip, pipx, or uv, grab the desktop app, or run it with Docker or Nix — then verify the install.Choose your surfaceTUI, desktop app, mobile relay, browser extension, exec mode, or the Python SDK — which surface to pick, and when.QuickstartSign in to a provider, launch the terminal UI, run your first task, and approve a tool call — the five-minute first-run path.Updating Local OperatorUpdate the runtime with lop update or /update, upgrade an install by hand, and find release notes.
By surface
Terminal (TUI)Launch, read, and drive the terminal UI — the transcript, the keys worth knowing, and the slash commands that matter.Desktop appInstall the desktop app, sign in, and learn the window — chat with run details, the tool trace, projects, schedules, themes, and the Agent hub inside the app.Mobile relayCheck on, steer, and start sessions from your phone — how the relay reaches your computer, how to set it up, and what the phone can do.Browser extensionGive the agent the browser you are already signed into — install from the Chrome Web Store, pair the bridge, and know exactly what it can and cannot do.
What it can do
Workflow automationsSchedule future turns, watch for changes, and run work in the background — the wake, monitor and job primitives, plus machine-readable output.Always-on & proactive agentsSessions that start their own turns — scheduled wakes that fire with every terminal closed, the supervisor that runs them, and Aida, the built-in chief of staff.Agent MeshPair your machines into one network so sessions can reach agents on other devices — roles, grants, onboarding, and incident controls.Agent HubBrowse Radient's community registry of agents, pull one into your setup, publish your own, and share privately with a team.
Build on it
Exec modeRun Local Operator headless — one-shot prompts, JSON event streams, background jobs, goal loops, and CI usage.Python SDKDrive sessions from Python — open or spawn one, deliver prompts, stream typed events, and collect results, with explicit roots and an approval policy.SSE streamingEmbed an agent runtime behind a Server-Sent Events stream — endpoint shape, event taxonomy, resuming with Last-Event-ID, keepalives, and working examples.What's in the ecosystemThe pieces around the runtime — model providers, Radient, MCP servers, skills, teams, the Agent Hub, and the browser extension.