> For the complete documentation index, see [llms.txt](https://docs.3techagency.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.3techagency.com/documentation/work-safely/access-privacy-and-safety.md).

# Access, privacy and safety

Role permissions, what is sent to your AI provider, retention, and the off-by-default safety switches.

## Access

Open **Settings > Access**. Administrators always have full access. Every other role starts with nothing.

| Column         | Grants                               |
| -------------- | ------------------------------------ |
| Chat           | AI Chat and the in-editor assistant  |
| Approve writes | Approving pending changes            |
| Run tasks      | Opening and running Tasks            |
| Woo-only       | Limits the role to WooCommerce tools |

Save the form for changes to apply. A grant never overrides WordPress's own object permissions or the editor lock. The assistant cannot grant itself access, and the model has no tool that changes these settings. See [Tools and permissions](/documentation/reference-and-help/tools-and-permissions.md).

**Protected pages**: WooCommerce cart, checkout and my-account and the privacy-policy page are always protected. Add more under **Settings > Advanced > Protected pages**. Writes to them are refused, not prompted.

## What is sent to your AI provider

When a model call is made, your messages, the page content the assistant reads, relevant Site Knowledge and tool results go to the provider configured in **Settings > Connectors**. The provider handles that data under its own terms. WP AI Manager adds a suggested paragraph to WordPress's privacy policy guide describing this.

Do not paste passwords, API keys or customer personal data into prompts or Site Knowledge.

## Retention and personal data

* **Settings > Privacy > Delete conversations after**: default **90 days**. A daily job deletes older conversations and Task runs. **0** keeps them forever.
* **Tools > Export Personal Data** includes a user's conversation transcripts. **Erase Personal Data** removes them.
* Agent-suggested memories stay **pending** until an administrator approves them in Site Knowledge > Memory.

## Off-by-default switches

| Switch (Settings > Advanced) | Why it is off                                                                                           |
| ---------------------------- | ------------------------------------------------------------------------------------------------------- |
| External MCP surface         | Lets outside AI apps control the site. In-admin Chat does not need it.                                  |
| AI can use advanced tools    | Filesystem, Database, WP-CLI, Custom Code and User-write tools can change PHP, database rows and users. |
| Allow AI-written code        | Loads activated PHP snippets. Each still needs a human to activate it in Sandbox.                       |
| Model choice in chat         | Keeps users on two curated tiers.                                                                       |

Keep them off unless you have a specific need and a recovery plan.

## Cost control

**Monthly token limit** (Settings > Advanced) is a soft ceiling. **0** means no limit. Past it, interactive Chat keeps working, but scheduled runs skip with a logged reason. The setting shows tokens used in the last 30 days. This is not a provider spending cap; set one in your provider account too.


---

# 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 by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.3techagency.com/documentation/work-safely/access-privacy-and-safety.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

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.