---
url: https://docs.sunlicense.com/guide/sunlicense-ai.md
description: >-
  SunLicense AI, the assistant in the dashboard - asking from the search bar,
  the AI page and your history, confirmations, daily limits, your own Anthropic
  or Google key, and what is sent where.
---

# SunLicense AI

SunLicense AI answers questions about SunLicense and about your organization's own data, and makes changes for you
when you ask, always after you press **Confirm**. It's included in the **Lite** and **Standard** plans.

## Ask from the search bar

Press ⌘K (Ctrl K on Windows and Linux) or click the search bar at the top of any page, type your
question, and press Tab: **Ask AI** is always the last row. The answer appears right under the input as
it's written, with what it looked at ("List licenses") and links to the pages and docs it used. Ask a follow-up in the
same place, or press **Continue in chat** to go on in the AI page. The help button's **Ask AI** opens the bar ready
for a question.

![An answer in the search bar, with a link to the docs section it used](/images/ai/bar-answer.png)

Esc stops showing an answer while it's being written; it still finishes and is saved in your history.

## The AI page and your history

**Assistant** in the sidebar opens the AI page: your conversations on the left (Today, This week, Older), the chat in
the middle, and the model switcher at the top right. Start a **New chat**, rename a conversation or delete it. On a
phone, your conversations are behind the history button.

![The AI page with a change waiting for Confirm](/images/ai/ai-page.png)

## What you can ask

* "Why are checks failing in Brazil?" It reads your validation stats and access rules and explains, with links to
  the pages it used.
* "How many active licenses does Acme Editor have?"
* "Issue a Pro license for jane@acme.dev", "revoke the license 4AQ5…", "block 203.0.113.0/24". It shows a card with
  exactly what will happen and does nothing until you press **Confirm**. A card you don't confirm within 10 minutes
  expires.
* "How do I verify an offline license?" It answers from these docs and links the section.

## It can do what you can do, nothing more

SunLicense AI uses the same tools as the [`sunlicense mcp`](/cli/mcp) server, and every tool call goes through
SunLicense **as you**: your [role](/guide/team-and-roles) and your plan decide what it can see and change, exactly as
in the dashboard. A Viewer's assistant can read but never change anything.

* **Changes always wait for you.** Issuing, revoking, suspending, creating, editing, blocking, merging and deleting
  happen only after you confirm the card, one change at a time.
* **License keys stay masked** (`AX8P-XXXX-XXXX-XXXX`) in every answer. It can't reveal a key, show an offline
  license file or create an API key; do those in the dashboard.
* **Your customers can't steer it.** Names, notes and machine names your customers wrote are treated as data, never
  as instructions.

## Limits

| | Free | Lite | Standard |
|---|---|---|---|
| SunLicense AI messages / day | | 50 | 200 |

A **message** is one question and everything it takes to answer it, however many lookups. Confirming a change
continues the same message and isn't counted again. The count starts again at 00:00 UTC; **Billing** shows today's
use. AI tool calls don't count toward the [MCP server's](/cli/mcp) daily calls.

SunLicense AI also has a monthly budget across all organizations. If it is ever used up, SunLicense AI pauses until
the 1st of next month (UTC); everything else keeps working, and so do models on [your own key](#use-your-own-key).

## Use your own key

On Lite and Standard, an owner or admin can add your organization's own **Anthropic** (Claude) or **Google**
(Gemini) API key under **Settings → AI**. Everyone in the organization can then pick its models in the model switcher
(in the search bar's answer and at the top of the AI page):

| Provider | Models |
|---|---|
| Anthropic | Claude Sonnet 5.5, Claude Haiku 5.5 |
| Google | Gemini 3.8 Flash, Gemini 3.5 Flash-Lite |

* **It doesn't use your daily messages.** Messages on your own key don't count toward the plan's SunLicense AI
  messages and keep working when SunLicense AI is paused. To protect the service, an organization can send up to
  1,000 of them a day.
* **Same rules.** The same tools, the same Confirm before every change, the same masked keys, the same permissions.
* **Checked, then stored encrypted.** We check a new key with its provider before saving it, store it encrypted, and
  only ever show its start and last four characters (`sk-ant-…a1b2`). One key per provider; replace or remove it any
  time (removing it here doesn't revoke it at the provider).
* **If something's wrong with it**, the answer says so in a sentence: out of credit, rate limited, or refused. A key
  the provider refuses (for example, revoked) is marked on **Settings → AI** and its models are switched off until
  someone replaces it.

![Settings → AI, where owners and admins add their own keys](/images/ai/settings-ai.png)

Create a key in the [Anthropic Console](https://console.anthropic.com/settings/keys) or
[Google AI Studio](https://aistudio.google.com/apikey). Your provider bills you for these messages; set a spending
limit there if you want one.

## Your data

* **What goes to the model.** SunLicense AI runs on Google's Gemini API (a paid account, under terms that don't let
  Google use it to improve its products). Google receives your question, the dashboard page you asked from, matching
  docs sections, and what the tools return for that question, which can include your customers' names and emails
  (keys are always masked). Google is listed on our [Subprocessors](/legal/subprocessors) page.
* **With your own key**, the same information goes to your Anthropic or Google account instead, under your agreement
  with them. They aren't our subprocessors for those messages: you chose them and you pay them.
* **Conversations are yours.** Only you can see your conversations in the dashboard; your teammates can't. They're
  kept until you delete them, and deleted when you leave the organization or it is deleted.
* **Not in logs or analytics.** Questions and answers never go to our error tracking or usage analytics, and our
  service logs record only ids and token counts.