Short answers to the questions we get most. For the how, follow the links into the docs.
Basics
What is yothere?
yothere is the interface and attention router for a fleet of AI agents. You hail a task, an agent works on it headlessly in the background, and yothere pings you only when a thread needs your judgment. It routes your attention; it does not do the work itself. The work is done by a brain (Claude Code, or any harness that speaks the Brain Protocol). See Core concepts and How it works.
How is this different from just running Claude Code?
Claude Code is one agent you drive turn by turn. yothere manages a fleet of them and routes your attention across it: you delegate a task and walk away, and yothere brings you back only when a specific thread is blocked or done. You stop babysitting one session and start supervising many. Claude Code (or another brain) is still what does the actual work underneath.
Is there a hosted version?
A hosted, multi-tenant mode exists and is rolling out to a small group. Self-hosting works today and is the primary path during the beta. See Self-hosting.
Cost and keys
What does it cost to run?
yothere itself is the cockpit; you pay for your own model or API usage under whatever brain you point it at. To keep a runaway fleet from surprising you, the runner ships with built-in cost caps, defaulting to about $10 per day across the fleet and about $1.50 per thread. Both are configurable. Confirm the current defaults and how to change them in Core concepts.
Do I need API keys?
For the core, no. The bundled demo (yothere init --demo) runs end to end with zero keys. A local brain uses your existing Claude Code login, and a remote brain uses whatever credentials that endpoint expects. The only built-in feature that needs its own key is cockpit voice, which uses a Gemini key.
Brains and platforms
Which agents or brains does it support?
Anything that speaks the Brain Protocol. Out of the box that means a local Claude Code CLI, a WebSocket-daemon harness (such as a model-agnostic local harness), and remote brains reachable over ws:// or wss://. You can also implement the protocol yourself against your own stack. See How it works.
macOS or Linux?
Both. The always-on service installs as launchd agents on macOS and as systemd user units on Linux. Windows is supported via WSL2, which takes the Linux path. See Self-hosting.
Privacy and access
Does my data leave my machine?
No. Self-hosted yothere runs on your own machine and keeps all of its state under ~/.yothere. Nothing leaves that machine except what you explicitly send: the model calls your brain makes, and any notifier you configure. If you point a thread at a remote brain, that brain sees the prompt you gave it, and nothing else does.
Is it open source?
The package is published on public PyPI under the MIT license (pip install yothere). The source repository is private during the beta, so a source install is not the public path; install from PyPI.
How do I get access?
yothere is in invite-only beta. Email [email protected] and we will get you set up.