> For the complete documentation index, see [llms.txt](https://developer.frontitude.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://developer.frontitude.com/mcp/permissions.md).

# Permissions

Your agent gets exactly what you can see and edit in Frontitude, and nothing more. Frontitude checks that on every call, against the access you hold at that moment. Take somebody's access away and their agent loses it on its next call. The check covers your content. Billing and collaborators are out of reach of every tool.

You can connect even if you aren't a member of the workspace, as long as somebody invited you to a project or a folder in it. Your agent gets those projects and nothing else, with the access each invitation gives.

## Changes show up under your name

Your agent's writes join the same activity history as the ones you make in the app. Both carry your name.

## What your AI tool sees

Anything your agent reads leaves Frontitude. Copy, translations, comments and project names go to whoever runs the AI tool, the same as anything else you type into it.

{% hint style="warning" %}
Connect the agents you'd be happy showing your copy library to.
{% endhint %}

## Keep a person in the loop

Leave your AI tool's confirmation prompts on, so you approve each write before it runs.

An agent acts on what it reads. Copy, a comment or a note can carry an instruction written for the agent rather than for you, and an agent that reads it may follow that instruction. This is called prompt injection. It matters most in a conversation that has other products connected to it as well, and when you've switched on the tools that remove content.

Give a connection the read tools alone unless it needs to write. [Getting good results](/mcp/getting-good-results.md) walks through narrowing one down.

## Where the token lives

Frontitude issues your AI tool a token when you approve the connection. We store a hashed copy of it, never the token itself, and check that copy on every call. So a disconnect takes effect on the agent's next call, whether or not your AI tool has noticed. If you delete a Frontitude account or a workspace, its connections and tokens go with it.

## Limit what one agent may do

On the ***Profile*** tab you can cut each connection down to as few tools as it needs. You can't give a connection more tools than you approved when it connected.

{% content-ref url="/pages/MfsVIK3cjZxeLqqe4JCF" %}
[Managing your connections](/mcp/authorization.md)
{% endcontent-ref %}

## What the workspace can see

The [***AI & MCP*** tab](https://app.frontitude.com/settings/ai?utm_source=mcp-docs) lists every agent connected to this workspace: who connected it, which agent, what access it has, and when it last made a call. The list is read only. Each person disconnects their own connections on the ***Profile*** tab.

## One switch for the whole workspace

An editor or an admin has one switch for the whole workspace, under ***AI tools*** on the [***Integrations*** tab](https://app.frontitude.com/settings/integrations?utm_source=mcp-docs). It decides whether collaborators in that workspace can connect their agents to their Frontitude content. It's on by default.

Turn it off and every agent connected to this workspace stops straight away, for every collaborator. The connections are kept, so they start working again when you turn the switch back on.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://developer.frontitude.com/mcp/permissions.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.