AI
AI settings
The AI settings section is used to manage AI models, assign models to tasks, add MCP servers, connect an AI search engine, enable a knowledge base. These settings are grouped into the corresponding tabs: AI models, Model assignment, MCP servers, AI search, Knowledge base. They are required for using the AI Chat and AI agents.
Activating AI features
To start using AI, activate the AI features add-on in one of the following ways:
- Click the Activate button in the banner at the top of the AI settings page.
- Go to Billing -> Add-ons and enable the AI features add-on.
AI features provide access to a wide range of AI models through the OpenRouter integration on a pay-as-you-go basis: you pay only for what you use, and the usage is billed from your Wallet (OpenRouter pricing + service fee applies to usage). If the available credits run too low, AI requests will not be processed.
On the Billing -> Add-ons -> AI features page, you can:
- enable or disable the AI features using the Enable AI features switcher,
- view the available credits and top up the Wallet,
- view the usage statistics: the spendings and the number of tokens processed for the current month,
- view the transaction history for the selected period, filtering it by a contact. The page displays the last 25 transactions; the full transaction history is available via the Download report button – the report is saved to Files.
Once the AI features are activated, the banner at the top of the AI settings page shows the AI features enabled status. Detailed transaction history, usage statistics, and service management are available via the Details button.
AI models settings
The AI models tab provides access to a wide range of AI models through the OpenRouter integration. Once the AI features are activated, the models are accessible to all users in AI chats.
The models are displayed as a list where each entry contains:
- the model name,
- the Input and Output prices per 1M tokens,
- the Off/On switcher which allows you to enable or disable the model,
- the Model details link which opens the model page.
Model assignment
The Model assignment tab lets you assign AI models to different tasks to customize performance and cost:
- Default AI setup – choose which model will be used by default in chats and agents.
- Custom models per task – use a different model for specific tasks instead of the default one: AI Chat, Code, Summarization, Translation, Text Analysis, Image generation, OCR, Vision, Form analysis. Only the models with the required capabilities are available in the list (for example, only image models can be selected for image generation).
MCP servers settings
This section lets you manage MCP servers for AI chats within rooms. The configuration is available to the workspace owner and Full admins only.
No MCP servers are available by default. The MCP servers tab contains the Edit configuration button only, and the configuration itself is empty:
{
"mcpServers": {}
}
The ONLYOFFICE Apps MCP Server, which streamlines AI-driven workflows by integrating LLMs with external systems for efficient and secure document management, is added manually, in the same way as any other server. To add it, follow the steps below:
- Click the Edit configuration button on the MCP servers tab.
- Add the onlyoffice-apps record to the mcpServers section, replacing the X-Mcp-Base-Url and X-Mcp-Api-Key values with your workspace URL and your personal API key generated in Developer Tools -> API keys:
{ "mcpServers": { "onlyoffice-apps": { "type": "http", "url": "https://mcp.onlyoffice.com/mcp", "headers": { "X-Mcp-Base-Url": "https://your-instance.onlyoffice.com", "X-Mcp-Api-Key": "your-api-key" } } } } - Click Save. The server appears in the Permissions section.
- Make sure that the onlyoffice-apps server is enabled.
The tools of the connected servers appear in the common tools list and are offered to the model. Both individual tools and entire servers can be enabled or disabled. If a server fails, an error icon with an expandable log is displayed in its row.
For detailed instructions on connecting the ONLYOFFICE Apps MCP server to this and other clients, refer to the MCP server documentation, the Connect to ONLYOFFICE Apps section.
Once enabled, the server tools are available to the other workspace users for their tasks, both in AI chats within rooms and folders and in AI agents, with the following limitations:
- In AI chat, the tools work with the access rights of the current user: a user can access only the rooms and files available to them.
- In an AI agent, only the members with the Agent manager or Content creator role can use the chat. The Viewer role has no access to it.
AI search settings
Use an AI search engine to enhance AI chats with real-time information from the internet. Web search and web crawling are powered by Exa. No Exa account or additional setup is required – the usage is billed through your workspace (Exa pricing + service fee applies to usage, pay-as-you-go from the Wallet).
The AI search tab displays the available services – Web search and Web crawling – with their prices per request.
To activate the service, click the Activate button in the banner at the top of the page, or enable the AI search add-on in Billing -> Add-ons. On the Billing -> Add-ons -> AI search page, you can enable or disable the service, top up the Wallet, and view the usage statistics and transaction history for the selected period.
Once configured, it will be available to all users in the workspace. The model can search the web, and the answer displays a clickable list of sources. AI search can be enabled or disabled at any time; when disabled, its tools disappear from the tools list.
Knowledge base settings
Use a knowledge base for AI agents. The system will index your documents using the text-embedding-3-small model to support question-based search through your data. This feature focuses on finding relevant information but doesn't generate summaries or perform global analysis.
The Knowledge base tab displays the Vectorization service with its price.
Once the AI settings are configured, you can start creating your AI agent. For more details on creating AI agents, please refer to our instructions.