FAQ

Is Vistara spying on me?

No. Vistara reads your agents' hook events and session logs on your Mac, plus git status for the folders they work in. The app has no screen-capture, clipboard-monitoring or trading code at all (not disabled: absent from the build), so it cannot see your screen or what you copy. The one clipboard read is opt-in: with Terminal typing on, it restores your clipboard after pasting a reply into Warp. It never reads files your agents do not touch. Browser tabs, Warp data and window titles are read only if you turn those features on. Before you turn on your node, only a daily anonymous "installed" ping leaves your Mac (a random install id, the app version and whether your node is on; you can turn it off). After that, only verification results: times, two pass/fail booleans and a hash. See Privacy and security.

Does it work with my terminal?

Yes. Vistara talks to Claude Code and Codex through their own hooks, and to Claude sessions through their own inbox sockets. Both work the same in Terminal, iTerm, Warp, or the terminal inside VS Code or Cursor. The only Warp-specific part is optional: typing a free-text reply into a Warp tab for an agent that has no other way to receive it.

Which agents does it support?

Claude Code fully: cards, answers, replies, live conversation and checks. Codex: cards and answers for approvals, live conversation and checks; free-text replies are typed into Warp if you allow it, otherwise copied. Cursor sessions are listed but cannot receive answers. See Agents and integrations.

Does it work with Claude Desktop?

Claude Desktop chats can ask Vistara about your agents: click ⚙ → Add to Claude Desktop, restart Claude Desktop, and ask "use vistara: what are my agents doing?". Desktop chats do not get cards, because Desktop's chat mode has no hooks. If Claude Desktop chat is the only AI tool you use, Vistara has little to show you yet: it tracks and answers coding agents running in a terminal. If you run Claude Code inside Claude Desktop, we expect it to use the same hooks as Claude Code in a terminal, but we have not verified that yet. See Agents and integrations.

Will Vistara approve things my agents shouldn't do?

By default, your local policy approves routine work inside the session's folder: reads, edits, tests, builds and similar commands. Everything else waits for you: destructive commands, anything outside the folder, secret files, network fetches to unlisted hosts, and changes to agent or Vistara settings. A text reply can never approve a permission. To approve nothing automatically, set "autoApprove": false in ~/.vistara-relay/policy.json. See The local policy.

What if I don't answer a card?

The agent waits up to 60 seconds, then shows its own prompt in the terminal, as it would without Vistara. Press Esc to hand it back sooner. If Vistara is closed, agents never wait on it.

What happens when my Mac sleeps?

Your agents pause when the Mac sleeps, with or without Vistara. Turn on ⚙ → Keep the Mac awake while agents work to stop idle sleep while an agent is mid-task. It releases as soon as nothing is working. The screen can still turn off, and closing the lid still sleeps the Mac.

How are points decided?

By the points ledger, on the server, from results your node signs. Today: +100 for each task a connected agent finishes (or work whose tests or build pass), +5 for each hour your node is online, and +20 for each UTC day on which any of your agents is active. Sessions and tokens earn nothing by themselves. Each result and its points appear under Rewards on the Points page. See Points and Vistara Node.

Is the node mining or staking?

No. The Vistara Node is the local daemon, a device key in your Keychain, and the signed results it sends. It does no mining, staking or validation, holds no funds, and runs no work for anyone else.

Do I need a crypto wallet?

No. Activating creates a wallet address on Base for rewards, with nothing to set up. The account is not deployed and holds nothing. Payments and the token are Planned and described only in the whitepaper.

Can I turn things off?

Yes:

• Reminder cards for a while: Shift+Esc on a card pauses reminders for sessions waiting on you for 15 minutes; Pause cards for 1 hour in the menu-bar menu does the same for an hour. Held approvals and questions still appear; press Esc on one to answer it in the terminal instead.
• Your agents' connection: ⚙ → Disconnect my agents removes Vistara's hooks from Claude Code and Codex.
• Optional features: browser tabs, Warp data and terminal typing are off unless you turn them on in ⚙ → Permissions.
• Auto-approval: "autoApprove": false in the policy file.
• Repo re-checks: "verify": false in the policy file.
• Everything: quit Vistara. Agents fall back to their own prompts.

What does it cost?

The beta is free. The whitepaper describes paid individual, team and enterprise tiers in the future (Section 5.4); pricing is not set.

Are there team features?

Not yet. Today you can hand work between agents on your Mac, or copy a handoff and paste it to a colleague. Planned for teams: presence, handoffs, questions routed to the right person, and shared verified results. Teams share state and results, not control of each other's machines. See the roadmap.

How do I report a problem?

Click ⚙ → Export diagnostics and send the file to mayur@vistara.dev, with a line on what you expected and what happened. See Getting started.

How do I delete my data?

Quit Vistara, remove its hooks (⚙ → Disconnect my agents), and delete ~/.vistara-relay/, ~/.vistara/, ~/Library/Application Support/@pouncer/shell/ and ~/Library/Logs/@pouncer/shell/. For records on the points server, email mayur@vistara.dev. Step-by-step instructions are in Privacy and security.

Why didn't a card appear for a session I already had open?

Agents load hooks when a session starts. Restart it once in the same tab with claude --resume <session-id> or codex resume <session-id>. With Show everything on, the overlay lists these sessions under Connect with the exact command.

How is this different from tools that run many agents in one window?

Those tools ask you to start agents inside them. Vistara works with agents wherever you already run them, and its focus is the state of the work and the moments that need you, not the window.