> For the complete documentation index, see [llms.txt](https://help.usejimo.com/docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://help.usejimo.com/docs/integrations/mcp/managing-access-usage-and-limits.md).

# Managing access, usage and limits

Revoke assistants, rotate API keys, change what an assistant may do, and follow the monthly allowance.

Everything lives on [Settings → Integrations → Model Context Protocol](https://i.usejimo.com/settings/integrations/mcp).

## Connected AI Clients

Each assistant you approved appears with its name and when it was last used. **Revoke** signs it out: it can no longer reach the project until someone connects it again.

<figure><img src="https://2794860263-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FKzAcDWQbK1gKbpra7bkb%2Fuploads%2Foh6XPNd0EtWI3K3Yx0ZQ%2Fimage.png?alt=media&amp;token=6854810e-8649-4da4-b2f8-7d21e2751dc2" alt=""><figcaption></figcaption></figure>

## API Keys

Each key shows its name, who created it, its Client ID, and, after a rotation, how long the previous secret remains valid. The menu offers **Rotate**, **Retire previous secret**, and **Revoke**. Details on [API Keys for Automations](/docs/integrations/mcp/api-keys-for-automations.md).

## Changing What an Assistant May Do

Connect the assistant again with the same URL. The approval page lists what it already has and what approving adds.

{% hint style="info" %}
New permissions reach existing assistants the same way. An assistant connected before October 7, 2026 gets **Build and run my Jimo Agent** only after you connect it again and approve it.
{% endhint %}

To take a permission away, revoke the client, then connect again with fewer boxes ticked.

## Usage & Limits

**Reading is free.** Overviews, funnels, feedback, agent test questions, and previews do not count. Reads are rate-limited to keep the service fair, never metered.

**Only executed changes count.** Your plan includes a monthly allowance of AI-driven changes, and a change is charged only once it has actually happened:

| Confirmed change                                       | Cost                                                  |
| ------------------------------------------------------ | ----------------------------------------------------- |
| Publish, pause, deletion, audience change              | 1 unit                                                |
| Draft or edit of an experience                         | 1 unit per executed change                            |
| Display rules, theme or position, each A/B test action | 1 unit                                                |
| Tags on an experience                                  | 1 unit per call, however many tags it adds or removes |
| Saved custom answer, agent setting change              | 1 unit                                                |
| Page, skill, or knowledge source created               | 1 unit each                                           |
| Translation                                            | 1 unit per language written                           |
| Document import                                        | 1 unit per document                                   |

A draft counts as one change, however many steps it has. A call that fails, is refused, changes nothing, or repeats an earlier identical call costs nothing. A batch the project cannot afford is refused whole at preview time and writes nothing.

**When the allowance is used up**, changes are refused with a "monthly change limit reached" message. Reading keeps working, and the allowance resets with your billing cycle.

{% hint style="info" %}
Every change result reports what it cost and what remains, so your assistant can warn you when you are getting close to the limit.
{% endhint %}

## Next Steps

* [Troubleshooting](/docs/integrations/mcp/troubleshooting.md): what the limit and permission messages look like, and what to do.
* [Permissions & Safety](/docs/integrations/mcp/permissions-and-safety.md): the full safety model.