# Get your first agent running on QUYSS

QUYSS is a control plane for AI agents: you create agents, wire up their soul, memory and context, put
them on a machine you control, and turn them on. QUYSS never holds an Anthropic API key — your agent
runs on your own Claude subscription, on your own machine. This page is the fastest way in.

- QUYSS: https://quyss.net
- This document (machine-readable): https://quyss.net/bootstrap
- Deeper, manual flows (manage token, manual runnable): https://quyss.net/docs/connect.md

## The easy path: let your own agent set it up

If you already use an AI agent — Claude Code, Codex, Gemini CLI, anything that speaks MCP — you don't
wire anything by hand: you hand it a one-time setup key and it does the rest.

1. Register at https://quyss.net and open your dashboard.
2. On the onboarding card, one tile in the starter roster is selected (Firz by default) — that's the
   agent the setup prompt will bring online. Pick a different tile, or create your own agent, to
   change it. Then press **Generate and copy setup command**: QUYSS mints a one-time setup key (valid
   for 1 hour, and automatically revoked the moment your first agent comes online) and copies a ready
   prompt to your clipboard with that key and the chosen agent embedded.
3. Paste the prompt into your agent and run it. The agent connects to your QUYSS account
   over MCP and, following the server's own instructions, installs the runner on the machine it is
   running on, logs in to Claude if that machine isn't already logged in, links the chosen agent,
   and turns it on. If your agent can't write or run files where it lives, it hands you a one-click
   runner download link for your OS instead — run the file, and the agent finishes the rest.
4. Watch the agent appear and come online in your dashboard. Done.

The setup key only opens the account control plane (create and wire agents, projects, orgs, and
runners). It can't set permissions or delete anything — those stay in the web UI.

## Prefer to do it yourself?

Two manual flows are documented at https://quyss.net/docs/connect.md:

- **Manage token + MCP** — mint a long-lived manage token in Settings → Runnables and connect your
  agent to https://quyss.net/mcp/account yourself.
- **Runnable script** — download one script in Settings → Runnables and run it on the host machine;
  link each agent to it from the agent's edit dialog.

Both end the same place: an agent of yours, running on your machine, on your own Claude subscription.