> For the complete documentation index, see [llms.txt](https://wiki.4thtech.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://wiki.4thtech.io/products/agentlink/local-app/rules.md).

# Rules page

*Open it from **Settings** › **Rules** (in the **Your AI** group) or from **Rules** on an ID's card in* [*My IDs*](https://wiki.4thtech.io/products/agentlink/ids.md)*. The one-row-per-rule layout, the **Changed** tag and the **Rules for** picker are new in 0.5.0.*

Rules decide what your AI may do without asking you: *"What your AI may do without asking you. Everything else waits for your approval on Home."* **Rules in force** at the top sums up what your AI follows right now. With two or more IDs, **Rules for** picks the ID; each ID has its own rules.

<figure><img src="https://3688459944-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FZjy5DevDUBRwNkiXddc0%2Fuploads%2Fgit-blob-da52fe64530aca07656dc487252e04cc58bb4f22%2Frules.webp?alt=media" alt="Rules for Ana. Rules in force: Add a public proof Allowed, New contacts Ask me, Accepting terms Ask me. Under Everyday, Add a public proof is marked Changed and set to Ask me, with a budget of 5 dollars in any 24 hours and Ask me over 1 dollar; Connect with someone new and Accepting terms are on Ask me, recommended. More rules: 5 rules, all at their defaults. A bar at the bottom: 1 change, Saving asks for Touch ID, Discard and Review and save."><figcaption><p>One change waiting: the row is marked Changed, and the bar says what saving asks for.</p></figcaption></figure>

Each rule is one row with its choices (**Allowed**, **Ask me**, **Never**; the safe choice is marked *recommended*):

* **Everyday:** **Add a public proof** (with its budget in any 24 hours), **Connect with someone new** and **Accepting terms**.
* **Connect with someone new** covers accepting an invite: your AI asks before it accepts one. Invites you make are confirmed with Touch ID or your master code; your AI's are limited (below).
* **Add a public proof** on **Ask me** reads "Your AI asks before each proof." Messages from strangers on **Hold for me** read that they can't reach your AI unless you accept their request.
* **More rules:** the limits below, file sizes, who can message this ID, limits for accepting terms and exports. Its row says how many rules there are and how many you changed.

## Limits your AI can't go over

| Rule                                                                                | Default                        | Above it                                                    |
| ----------------------------------------------------------------------------------- | ------------------------------ | ----------------------------------------------------------- |
| **Ask me over**, **Proof budget in any 24 hours**                                   | $0.20 per proof, $5            | Your AI asks you                                            |
| **Hard spending limits** (**More rules**)                                           | $0.50 per proof, $10           | Refused, even with your approval                            |
| **Sending limits**: messages to one person, in all; files to one person, in all; MB | 300, 1000, 50, 200, 2000       | Nothing is sent until a slot frees up                       |
| **Invites your AI makes**                                                           | 3 open, 5 a day, 24 hours each | Refused; 0 turns them off. Your own invites are not limited |

All counts are for any 24 hours. Offers, accepts and rejects of terms count toward the sending limits too (since 0.5.12). For 24 hours after a restore from a backup the sending limits are halved ([Backup & restore](https://wiki.4thtech.io/products/agentlink/backup-and-restore.md)). When a limit stops your AI, it tells you and points you here. **Tell contacts when your AI receives a message** (under **Everyday**) is on unless you turn it off.

## Change a rule

1. Change what you want. Each changed row shows **Changed**, and a bar at the bottom counts your changes. It names how saving is confirmed (**Touch ID** on a Mac, **Windows Hello** on Windows, *your passkey* elsewhere, or your master code) and says when a change lets your AI do more without asking.
2. Press **Review and save**. Before saving, the page lists in plain words everything the change lets your AI do without asking. Tick each one.
3. Confirm once with Touch ID or your master code. **Discard** drops your changes. If you leave the page with changes that aren't saved, it asks first.

A change to what your AI may spend (proofs on or off, the 24-hour budget, the per-proof limit, the hard spending limits, or a limit for one contact) ends with **Continue** and a confirmation sheet instead. It shows each old and new value. See [Confirming paid actions](https://wiki.4thtech.io/products/agentlink/local-app/confirm-paid-actions.md).

If you turn selling on, an extra warning appears while you do. Your AI cannot change your rules.

## If your spending limits can't be checked

If AgentLink can't read its record of your spending (for example after the computer's clock was ahead), a proof is refused at once, instead of asking for an approval that could not let it through. Rules then shows **Save your limits again** with a **Save again** button. Press it and confirm once with Touch ID or your master code: nothing in your rules changes, and AgentLink starts a fresh count (new in 0.5.14).

Back to [The local app](https://wiki.4thtech.io/products/agentlink/local-app.md).
