Queen FAQ
Short answers to the common questions about Queen. Queen is in active design; answers describe what it does by design.
Queen FAQ
Short answers to the common questions about Queen. Queen is in active design; answers describe what it does by design.
Related:
#Can Queen read my memory, prompts, or session text?
No. The control plane holds coordination state and coarse usage facts only. It never holds memory content, prompts, completions, session text, tool-call payloads, file paths, repo names, or plaintext credentials. There is no column that could hold session text.
#Who holds the credential that unlocks my team's shared memory?
A durable machine in your own trust domain, a laptop custodian or fleet orchestrator, holds custody of the memory credential, not the cloud. Queen coordinates around data it cannot open in the default mode.
#How do I add a second machine to my fleet?
Mint a short-lived join token on an already-trusted machine, then redeem it on the new one. No browser on the server, no credential pasted into a config file, no shared key. The token joins the machine and expires; the machine earns its own identity and starts reporting in.
#What happens if a machine is stolen or an engineer leaves?
Two honest, separate steps: revoke the device in Queen, a control-plane action, then rotate the memory credential, a data-plane action. Per-agent identity means one revocation cuts off one agent, not the fleet.
#What if the command authority goes down?
The fleet keeps working. Agents cannot receive new commands but keep doing local work and keep reporting in. They degrade to autonomous, not dead. The authority is needed to issue commands, never to run workers.
#Can Queen run offline or air-gapped?
No. Binding to a cloud deployment is a required step, fused with license enforcement. There is a bounded grace window for a transient authority outage only; anything else fails closed with a clear reason.
#Does Queen intercept my desktop apps?
Not in the first version. Observation is limited to command-line and model-gateway traffic, and to coarse facts only. Desktop interception would require installing a certificate authority on every machine and is explicitly deferred, off by default, behind a future consent-gated design.
#Why is there no local web dashboard on the agent?
By design. Any dashboard renders fleet, org, or ROI data, which belongs behind the cloud application's sign-in, not on a loopback port on a possibly throwaway machine. The local agent is command-line only.
#How is per-user ROI kept honest and private?
Per-user leaderboards stay inert until a verified backend identity exists, so there are no self-asserted names or fabricated rows. Per-user spend is treated as personal data, with visibility controls and an erasure path. Every cost figure carries its basis, measured or allocated, and the two are never blended silently.