Dashboard Walkthrough
The hosted dashboard at nfltr.xyz/dashboard shows what the relay knows: which of your hubs are connected, the machines agents can run on, and your keys and settings. Runs stay with your hub on its own machine, end to end encrypted; the dashboard points there. Every view has a CLI cross-check.
Open nfltr.xyz/dashboard and sign in with the account whose API key your nodes and hub use. The dashboard reads no hub content; the hub decides. To steer, continue, or stop a hub agent, ask the hub in its session (it calls send_message or stop_agent).
1. Check readiness first
$ nfltr orch doctor
$ nfltr orch nodes list
$ nfltr orch fleet workers --jsondoctor checks the Claude CLI and its auth, the relay, and connected nodes or workers. nodes list shows the machines a hub can spawn agents on; see Join machines if it is empty.
2. The four views
The bar at the top switches between four views. Each has its own link (#agents, #machines, #runs, #account), so you can bookmark or share one.
| View | What belongs there | CLI cross-check |
|---|---|---|
| Agents | Your hubs as the relay sees them: online or offline, last seen, and the reattach command for a stopped one. No hub content | nfltr orch hub runs |
| Machines | Joined machines, their agent slots and what each can run (harnesses, monitors), the agent catalog with terminals, and worker coverage | nfltr orch nodes list |
| Runs | A pointer only: where your runs are and the commands that show them; it makes no request | nfltr orch hub show --hub-id ID AGENT_ID |
| Account | API keys and your settings | nfltr orch doctor |
Hubs stay end to end encrypted: a hub sends the relay no summary or status of its runs, so nfltr.xyz never sees a run. See What the relay sees.
Platform admins also see Admin: cluster metrics, fleet overview, fleet tokens, global settings and contact messages.
A stopped hub shows as offline with its reattach command, nfltr orch --hub-id ID; its agents may still be running on their machines, and the reattached hub receives every completion it missed.
3. Give the hub a goal and watch it land
Start work from a terminal or your own Claude Code session (see Give the hub a goal; Start a hub on the Agents view shows the same commands), then follow its agents in the hub session, or list them on its machine:
$ nfltr orch "add a regression test for the cache bug and open a branch"
$ nfltr orch hub runs
$ nfltr orch task list --active4. Read a task and its result
On the hub's machine, running or not, nfltr orch hub show prints one agent's turns: each message, state, worker, usage, error, branch and sha, the answer and result, and the last events of the turn's task. While the hub runs, nfltr orch task reads the same task from any machine of the account, end to end encrypted.
$ nfltr orch hub show --hub-id HUB_ID AGENT_ID --events 20
$ nfltr orch task status --task TASK_ID --events 20
$ nfltr orch task result --task TASK_IDFor a hub agent, task status shows a Hub: line naming the hub that owns it.
5. Answer questions and approvals
The hub asks you in its session; a hub run with --remote-messages=answers also reaches a device paired with it (nfltr orch hub pair) on the dashboard's Hubs page, end to end encrypted. While the hub runs, answer a task's question from a terminal:
$ nfltr orch task answer --task TASK_ID --message "your answer"Review an approval's request and objective (nfltr orch task status), then approve or reject it:
$ nfltr orch task approve --task TASK_ID --reason "evidence reviewed"
$ nfltr orch task reject --task TASK_ID --reason "evidence is incomplete"6. Know which controls change state
Refresh, opening cards, and copying commands are read-only. Deleting agent records and changing keys or fleet tokens affect external state; check the target before you press them. The dashboard has no task controls: hub agents are steered and stopped through their hub.
Next: Observe orchestration for reading runs in the hub and Team collaboration for sharing a run.