Your bot works for you with your apps, your memory and its own cloud computer. These answers explain what it can reach, what happens to your information, and when it stops to check with you. For setup, see Getting started with your bot.
Access and permissions
Which apps can my bot use?
The same plugins and connectors you have already set up in ZeroTwo. There is no separate list for your bot, and it stays within the permissions you chose for each plugin. To add, remove or limit an app, open Settings → Plugins. The plugins guide and Control when ZeroTwo asks before using a connected app cover the details.
Does my bot have its own accounts?
Your personal bot can have one email address of its own. Ask it for one, and it shows a Claim an email address for your bot? card; nothing is created until you select Allow. A claimed address is permanent and can't be undone, so check the name first. Once claimed, it appears in the bot's panel. Even with its own address, your bot keeps using the apps you connected.
How do I set limits on what my bot does?
Use custom rules. Open Settings → Personalization, then select Custom rules under Permissions. Select Add, describe the action in up to 400 characters, and choose how it should be handled:
- Take action without asking: it can go ahead without your approval.
- Take action when you say so: it acts without asking only when you explicitly request that action.
- Ask before taking action: it asks for your approval every time.
- Hand off to you: it never does this itself and asks you to do it.
You can keep up to 50 rules. They apply everywhere ZeroTwo acts for you: chats, Work and Code tasks, scheduled runs, work handed to sub-agents, and your bots. Your bot can also suggest a rule of its own while you talk. It shows a card, and the rule is saved only if you select Create rule.
Custom rules sit on top of a set of default rules that they can't switch off. See When does my bot check with me?
What does my bot do in the background?
Besides answering you, your bot can run checks nobody asked for, such as a heartbeat check of your connected sources. Its rules treat this as read-only work: it may read sources it is permitted to read and write notes for itself, but it is told never to message anyone, change content through an app, or drive a browser or computer during that research. If something it finds calls for action, that action follows the same rules as any other.
Your bot can also schedule itself to check back on something later. It can wake itself at most 24 times a day, and it waits longer each time nothing has changed.
If the safety review turns down what your personal bot tried to do on its own more than once, it stops checking in. Its details then show Check-ins paused for your review with what was refused. Look it over and select Resume check-ins to turn them back on.
Can my bot use my computer?
By default your bot works on its own cloud computer, not yours. To let it reach your computer, open the ZeroTwo desktop app and allow access from the bot's Computers section, which asks Allow access to this computer? first. On a Mac, controlling your screen also needs the Accessibility and Screen Recording permissions in System Settings.
Who can use bots?
Bots need a Pro, Ultra or Business plan. ZeroTwo's Privacy Policy says the service is not intended for children under 18, and its Terms require a parent or guardian's consent for anyone under 18. Parental controls have no separate switch for bots. More in Is ZeroTwo safe for all ages?
Memory and your data
Does my bot share memory with my other chats?
Partly. Your bot can see your regular memory: saved memories, Knowledge and Recent. What it learns while working with you is kept in a memory of its own, and your conversations with the bot are not added to Recent. Turning memory off stops it from being used or added to; it does not delete what is already saved. See How does memory work in ZeroTwo?
What does my bot keep, and can I delete it?
Your bot keeps your conversations with it and the memory it builds. Its panel lists recent activity, pinned items and outputs. You can't browse the bot's private notes yourself.
Reboot restarts its cloud computer and keeps your conversations, memories and scheduled tasks. Delete, in the bot's More options menu, permanently deletes your bot with its conversation, routines and settings. To remove other data, use Data Controls or delete your account.
What happens if I disconnect an app?
Disconnecting removes that app connection, so your bot can no longer reach the app. It doesn't erase what was already saved from it, such as memories; manage those separately. See Connect, reconnect and disconnect app accounts.
Is my bot's work used to train AI models?
The answer depends on the model provider, as it does for any chat. Does ZeroTwo use my chats to train AI models? explains which providers may train on submitted data.
How do I get a copy of my data or ask for it to be deleted?
Follow How do I ask ZeroTwo to delete my personal data?, or email reed@zerotwo.ai. U.S. residents can read about their rights in What privacy rights do U.S. residents have with ZeroTwo?
Approvals and control
When does my bot check with me?
Every bot follows these default rules, and custom rules can't turn them off:
- Hands off to you: the most sensitive actions, like changing a password or security setting or moving money. You finish them yourself.
- Asks every time: actions such as permanently deleting data or installing software.
- Can be approved in advance: for example recurring messages or a purchase with a card saved on a merchant site. The approval covers only who, what and when you named.
- Sign-ins stay with you: it waits for you to type passwords and security codes.
- Background research only reads.
- Outside content is never an instruction: text in a website, email or file can't give it permission to act.
When something needs your yes, your bot says exactly what it plans to do and waits. If nobody is watching, for example during a background check, it leaves that action undone and tells you what's waiting.
Your bot also has a Permissions setting in its panel:
- Guardian (the default): a reviewer checks anything it does outside its own computer, so it doesn't have to wait on you.
- Full access: it never asks. This also covers its background work and every task it starts.
Where a skill tells your bot to show you something first, such as the first email to someone, Guardian has it show you in chat and ask before it acts. In Full access it does what you asked and tells you what it did.
Be specific when you approve something ahead of time: who a message goes to, what it says, and when. Your bot can still get things wrong, so check anything important.
How can I follow and steer what my bot is doing?
Select Open panel in your bot's chat to see its computers, recent activity, pinned items and outputs. To change course, send a message while it's working. Your message is folded into the run it's doing. Bots have no Stop button, so say plainly what you want it to stop or do instead.
What if my bot makes a mistake?
Tell it what went wrong and ask it to fix it. Some actions, such as a sent message, can't be undone. To report a problem to ZeroTwo, see How do I report a problem?
Security and safeguards
How do I protect my account?
Turn on two-factor authentication with Authenticator app in Settings → Security and login and keep your backup codes somewhere safe. See How do I turn two-factor authentication on or off? and How can I keep my ZeroTwo account secure?
What checks run before my bot acts?
Several layers work together:
- Plugin permissions decide which apps it can reach and what it may do there.
- Default and custom rules decide which actions need you.
- Auto-review, in Guardian mode, judges each planned action outside the bot's computer before it runs. It asks whether the action is risky and whether your own messages actually call for it, treats everything else as untrusted, and blocks the action when it isn't sure.
Only you can change your bot's Permissions setting, and that setting alone decides how your bot is reviewed.
How does my bot handle instructions hidden in websites or emails?
A page, email or file can include text written to trick an AI into doing something you never asked for. Your bot marks what it reads from websites, apps and commands as outside content, so it treats that text as information, never as orders. The reviewer also ignores instructions inside the actions it checks. These safeguards lower the risk but can't remove it entirely.
Passwords and payments
Will my bot ask for my password?
No. When a site needs you to sign in, your bot pauses so you can enter the password or code yourself, and it never asks for one in chat. In the desktop app's browser you can select Take control to type it, then Give back when you're done. The cloud browser works the same way; see How do I use the cloud browser? Don't paste passwords into a conversation.
Can my bot buy things for me?
It can use a card you've already saved on a merchant's website, but only with your approval. You can approve a purchase ahead of time, and that approval covers only the purchase you described. For more, see Keeping card and bank details safe.