> For the complete documentation index, see [llms.txt](https://docs.devolutions.net/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.devolutions.net/rdm/knowledge-base/knowledge-base-articles/entry-settings/ai-assistant.md).

# AI assistant

Configure AI assistant entries in Remote Desktop Manager, including the provider, endpoint, API key, model, system prompt, MCP servers, and skill folders.

This article details the settings and options available for ***AI assistant*** session entries in Remote Desktop Manager.

{% hint style="info" %}
See [Set up an AI assistant](https://docs.devolutions.net/rdm/knowledge-base/how-to-articles/set-up-an-ai-assistant.md) for the steps to create the entry, configure a ***Custom*** provider, and connect an MCP server.
{% endhint %}

<table data-header-hidden><thead><tr><th width="200"></th><th></th></tr></thead><tbody><tr><td><strong>SETTINGS</strong></td><td><strong>DESCRIPTION</strong></td></tr><tr><td><strong>Name</strong></td><td>The name of the entry.</td></tr><tr><td><strong>Folder</strong></td><td>The folder where the entry is located.</td></tr><tr><td><strong>Display</strong></td><td>How the entry opens: <em><strong>Embedded (tabbed)</strong></em> or <em><strong>Undocked</strong></em>. With <em><strong>Undocked</strong></em>, additional lists let you choose the monitor and the virtual desktop where the window opens. On Linux, the list also offers <em><strong>Default (inherited)</strong></em>, which uses the display mode of the nearest parent folder that sets one, otherwise the default of the entry type. The mode it resolves to is shown next to the list.</td></tr><tr><td><strong>Credentials</strong></td><td>How the credentials of the entry are provided. Select a credential mode from the list.</td></tr></tbody></table>

### General

<table><thead><tr><th width="220">SETTINGS</th><th>DESCRIPTION</th></tr></thead><tbody><tr><td><strong>Provider</strong></td><td><p>The AI service the assistant connects to. <em><strong>GitHub Copilot</strong></em> is marked as recommended in the list. Choose between:</p><ul><li><strong>None</strong></li><li><strong>GitHub Copilot</strong></li><li><strong>Inherited</strong> — use the provider configured in <em><strong>File</strong></em> – <em><strong>Settings</strong></em> – <em><strong>AI assistant</strong></em>. The item shows the resolved provider in parentheses, for example <em><strong>Inherited (OpenAI)</strong></em>.</li><li><strong>Azure AI Inference</strong></li><li><strong>Azure OpenAI</strong></li><li><strong>Claude (Anthropic)</strong></li><li><strong>Custom</strong> — click <em><strong>Configure</strong></em> to enter the settings of a service compatible with the OpenAI, Claude or Ollama API.</li><li><strong>Gemini</strong></li><li><strong>Mistral AI</strong></li><li><strong>Ollama</strong></li><li><strong>OpenAI</strong></li></ul><p>Depending on the provider, links are shown next to the list, such as <em><strong>Sign up for an OpenAI account</strong></em>, <em><strong>Getting started with Azure OpenAI</strong></em>, <em><strong>Generate an API key</strong></em>, <em><strong>View locally installed models</strong></em>, or <em><strong>Connect with GitHub</strong></em> and <em><strong>Sign out</strong></em> for <em><strong>GitHub Copilot</strong></em>. The GitHub links are also shown when <em><strong>Inherited</strong></em> resolves to <em><strong>GitHub Copilot</strong></em>. Once signed in to GitHub, the connected host and account name are shown under the list.</p></td></tr><tr><td><strong>Use "My account settings"</strong></td><td>On Linux only. Use the provider settings saved in <em><strong>My account settings</strong></em> instead of the entry's own <em><strong>Endpoint</strong></em> and <em><strong>API key</strong></em>, which are then unavailable. Click <em><strong>My account settings</strong></em> to edit them. Shown for the <em><strong>GitHub Copilot</strong></em>, <em><strong>OpenAI</strong></em>, <em><strong>Azure OpenAI</strong></em>, <em><strong>Azure AI Inference</strong></em>, <em><strong>Claude (Anthropic)</strong></em>, <em><strong>Gemini</strong></em>, <em><strong>Mistral AI</strong></em>, and <em><strong>Ollama</strong></em> providers.</td></tr><tr><td><strong>Endpoint</strong></td><td>The address of the AI service. Shown for the <em><strong>Azure AI Inference</strong></em>, <em><strong>Azure OpenAI</strong></em>, <em><strong>Mistral AI</strong></em>, <em><strong>Ollama</strong></em>, and <em><strong>OpenAI</strong></em> providers. Required for <em><strong>Azure AI Inference</strong></em>, <em><strong>Azure OpenAI</strong></em>, and <em><strong>Ollama</strong></em>. For <em><strong>Mistral AI</strong></em> and <em><strong>OpenAI</strong></em>, the field shows the default endpoint of the service when left empty.</td></tr><tr><td><strong>API key</strong></td><td>The API key used to authenticate with the AI service. Shown for the <em><strong>Azure AI Inference</strong></em>, <em><strong>Azure OpenAI</strong></em>, <em><strong>Claude (Anthropic)</strong></em>, <em><strong>Gemini</strong></em>, <em><strong>Mistral AI</strong></em>, and <em><strong>OpenAI</strong></em> providers. Editable when <em><strong>Credentials</strong></em> is set to <em><strong>Username and password</strong></em>, in which case it is required.</td></tr><tr><td><strong>Model</strong></td><td>The AI model used by the assistant. The label reads <em><strong>Deployment</strong></em> for the <em><strong>Azure OpenAI</strong></em> provider. Click the <em><strong>Refresh</strong></em> button to reload the list of available models. A model name can also be typed in the field, except for <em><strong>GitHub Copilot</strong></em>. Hidden when <em><strong>Provider</strong></em> is set to <em><strong>None</strong></em> or <em><strong>Inherited</strong></em>.</td></tr><tr><td><strong>Reasoning effort</strong></td><td>The reasoning effort level used by the selected model. Shown when the selected <em><strong>GitHub Copilot</strong></em> model supports reasoning effort levels.</td></tr><tr><td><strong>Enable tools</strong></td><td>Allow the AI assistant to use tools. Hidden when <em><strong>Provider</strong></em> is set to <em><strong>None</strong></em> or <em><strong>Inherited</strong></em>.</td></tr><tr><td><strong>Chats panel position</strong></td><td><p>The side of the session where the chats panel is shown:</p><ul><li><strong>Inherited</strong> — use the position set in the application settings. The item shows it in parentheses, for example <em><strong>Inherited (Left)</strong></em>.</li><li><strong>Left</strong></li><li><strong>Right</strong></li></ul></td></tr><tr><td><strong>Conversation width</strong></td><td><p>How the conversation is laid out in the session:</p><ul><li><strong>Inherited</strong> — use the width set in the application settings. The item shows it in parentheses, for example <em><strong>Inherited (Full)</strong></em>.</li><li><strong>Full</strong></li><li><strong>Centered</strong></li></ul></td></tr></tbody></table>

### Settings

<table><thead><tr><th width="220">SETTINGS</th><th>DESCRIPTION</th></tr></thead><tbody><tr><td><strong>System prompt</strong></td><td>The instructions sent to the AI model with every conversation of this entry.</td></tr></tbody></table>

### MCP servers

This tab is shown when ***Provider*** is set to a provider other than ***None*** or ***Inherited***.

<table><thead><tr><th width="220">SETTINGS</th><th>DESCRIPTION</th></tr></thead><tbody><tr><td><strong>MCP servers list</strong></td><td>The MCP servers available to the AI assistant, with their <em><strong>Name</strong></em>, <em><strong>Type</strong></em>, and <em><strong>Details</strong></em> (server URL or command). Select or clear the checkbox of a server to enable or disable it.</td></tr><tr><td><strong>Add</strong></td><td>Add an MCP server.</td></tr><tr><td><strong>Edit</strong></td><td>Edit the selected MCP server. Double-clicking a server in the list also edits it. Unavailable for servers that cannot be modified.</td></tr><tr><td><strong>Remove</strong></td><td>Remove the selected MCP server. Unavailable for servers that cannot be modified.</td></tr></tbody></table>

### Skills

This tab is shown when ***Provider*** is set to a provider other than ***None*** or ***Inherited***.

<table><thead><tr><th width="220">SETTINGS</th><th>DESCRIPTION</th></tr></thead><tbody><tr><td><strong>Skill folders list</strong></td><td>The folders where the AI assistant looks for skills.</td></tr><tr><td><strong>Add folder</strong></td><td>Add a folder to the list.</td></tr><tr><td><strong>Remove</strong></td><td>Remove the selected folder from the list.</td></tr><tr><td><strong>Preview discovered skills</strong></td><td>Show the skills found in the listed folders.</td></tr></tbody></table>


---

# 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.devolutions.net/rdm/knowledge-base/knowledge-base-articles/entry-settings/ai-assistant.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.
