> ## Documentation Index
> Fetch the complete documentation index at: https://docs.guardway.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Assistant

> The in-dashboard AI agent — ask about your Guardway organization and take governed actions, with an approval step before anything is weakened.

## What this is for

**Guardway Assistant** is an AI agent built into the dashboard. Ask it about your organization — spend, policies, keys, anomalies, the supply chain, the endpoint fleet — and have it take actions for you: the same governed capabilities the [Claude connector](/platform/claude-connector) exposes, without leaving Guardway. Each conversation is kept as a thread you can reopen.

> [!NOTE]
>
> Assistant has **limited availability**. It appears only when the `assistant` feature is enabled for your organization and you're an **Owner** or **Admin**. Ask your Guardway contact to turn it on.

## Options

* **Ask about your org** — usage and spend, effective policy, API keys, alerts and anomalies, the supply chain, and the endpoint fleet.
* **Take actions** — create or change keys, set budgets, adjust policy, acknowledge anomalies and more, each run **as you**, with your role.
* **Approval before weakening.** A change that loosens protection (widening a policy, raising a budget) surfaces an **approval card** — **Approve** or **Deny** — before it runs. Nothing is ever deleted from chat.
* **Threads.** Each conversation lives at `/assistant/<thread>`; a non-owner admin opens an existing thread read-only.

## How to configure

**1. Enable it**

An Owner turns on the `assistant` feature for the organization (ask your Guardway contact if you don't see it).

**2. Open the Assistant**

Select **Assistant** in the sidebar, or use the header launcher. Ask a question or request a change.

**3. Review actions**

When the Assistant proposes a change that weakens protection, review the **approval card** and **Approve** or **Deny** it.

## Permissions

The Assistant acts as you, with your role: **Owner** and **Admin** can run actions; other roles — and a non-owner admin on someone else's thread — are read-only. It never deletes from chat, and secrets never pass through it.

## Related

* [Claude Connector](/platform/claude-connector) — the same governance from Claude Code and the Claude apps, outside the dashboard.
* [Agents](/platform/agents) — the AI agents Guardway discovers across your providers (distinct from this in-dashboard assistant).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.