FAQ

Claude, Codex, Antigravity, Grok, Cursor, GLM, MiniMax, WorkBuddy and Kimi. Every one shows sessions and status; how deep quota and returning to a session go differs by Agent, and the app says so.

No. Most people meet it through coding Agents, but the questions are the same for any Agent: what is working, whose turn it is, why it is slow, and what it cost.

It reads local evidence on this computer: session files, processes and logs. When the evidence is not enough it shows No status. It does not guess.

Working, Your turn, Stuck, Waiting for service, Waiting for tool, Waiting for MCP, Turn complete and No status. At the provider level there are also Quota limited and Sign-in required. One row per session, and several sessions under the same Agent each get their own row.

Top bar, sidebar and floating ball. The top bar sits at the top of the screen, as a notch or a small floating bar just under the top edge, your choice; the floating bar stays hidden until the pointer moves over it. The sidebar docks on the left or right edge, rests as a small notch and expands inward as the pointer approaches. The floating ball is a small ball with the Agent that needs you most at its center, and you can drag it anywhere. All three are in Free; glass, display style and shimmer are Pro.

The deep network check goes link by link and tells you where it breaks and how to fix it; once fixed, you can check again. looptend does not change your proxy, DNS or VPN.

Free gives a one-line verdict: Stable, Near load, Clearly limited or Severely overloaded. Pro’s deep performance check finds the processes slowing your computer down, such as helper processes an Agent left behind; whether to end them is your call, and it rechecks before ending anything.

Loop is the cycle between an Agent working on its own and the moment it comes back to you. Tend means watching over it and keeping it turning. looptend keeps every loop visible and calls you only when it is your turn. The full story is here.

No. It puts the request in front of you and “Open session” takes you back to that session. Approvals change files, permissions and accounts, so that step stays with you.

looptend Work: what each person’s Agents finished is written into daily and weekly reports once that person confirms it, and the team aligns on work that was actually done. Priced per seat.

Free shows the quota of every account. With Pro and “Switch automatically when quota runs out” turned on, once an account is confirmed used up, the next session automatically uses an account that still has quota.

Free: sessions and status for nine Agents, your-turn alerts and “Open session”, live quota, network and performance monitoring, the usage report and weekly and monthly report images. Pro: everything in Free, accounts with automatic switching when quota runs out, the deep network check and deep performance check, the work summary with Word and PDF export, and appearance customization. Max: everything in Pro, five or more accounts, several computers, and Agents running unattended for long stretches. Prices are under Pricing.

No. looptend switches accounts only for the next session and never migrates a running thread. A session that is already running stays on its account until it ends.

No. An unreadable quota, a stale reading, an expired login or a network failure never counts as used up. After several failed reads it only notes that the quota is temporarily unreadable and keeps the last reading; it switches only on verified evidence that the quota is gone.

Accounts of the same Agent: one Codex account runs out, the next session uses another Codex account. Kimi, GLM and MiniMax plans can be set up too; Claude quota is read only, never switched. Each Agent keeps its own quota. The full explanation is here.

Subscription mode reads the quota and status behind your official sign-ins with each vendor. API mode takes an address and a key; looptend recognizes an official API or a relay endpoint, lists the models and reads the balance. Automatic account switching works in subscription mode only.

No. Logins always stay with the official process; looptend only reads quota and status.

Status, quota and diagnostics are all computed on your computer. Live network monitoring sends only empty requests with no content, never prompts, session text or code. Only the work summary, when you press “Generate”, sends excerpts with secrets removed to looptend’s summary service, and nothing is stored.

No. The precise speed test, the deep network check and the deep performance check run only after you start them, and stop when they finish.

Within the time range you chose: the first prompt and last reply of each session, plus the project’s git commits. With secrets removed they go to looptend’s summary service to generate the summary, then they are discarded and never saved to disk. You do not configure an API. The full explanation is here.

Reports are generated on your computer and sent only after you confirm. No scores, no rankings, and the raw record never leaves your machine.