# Getting started with AI

Comind.work integrates AI capabilities that let you query, analyze, and interact with your workspace data using natural language.

## AI capabilities overview[​](#ai-capabilities-overview "Direct link to AI capabilities overview")

| Feature                                              | Description                                                                     | Status         |
| ---------------------------------------------------- | ------------------------------------------------------------------------------- | -------------- |
| [MCP tools](/api-integrations/ai/mcp-tools.md)       | Connect external AI assistants (Claude, ChatGPT, etc.) to your Comind.work data | Available      |
| [AI Assistant](/api-integrations/ai/ai-assistant.md) | In-app AI chat for workspace queries                                            | In development |

## MCP - the fastest way to start[​](#mcp---the-fastest-way-to-start "Direct link to MCP - the fastest way to start")

The Model Context Protocol (MCP) connects your Comind.work data to any AI tool that supports MCP. This is the recommended way to start.

### What you can do with MCP[​](#what-you-can-do-with-mcp "Direct link to What you can do with MCP")

* **Query data** - "Show me all open tickets assigned to me"
* **Aggregate and analyze** - "How many tasks were completed last month, broken down by team?"
* **Search** - "Find records mentioning 'budget approval'"
* **View history** - "What changes were made to TASK123 last week?"
* **Read attachments** - AI can read images and documents attached to records

### Quick setup[​](#quick-setup "Direct link to Quick setup")

1. Get your MCP server URL and authentication token from your Comind.work administrator

2. Add the MCP server to your AI tool:

   <!-- -->

   * **Claude Desktop** - add to `claude_desktop_config.json`
   * **Claude Code** - add to `.mcp.json`
   * **VS Code** - configure in MCP extension settings

3. Start asking questions about your data

See [MCP tools](/api-integrations/ai/mcp-tools.md) for the detailed setup guide.

## How AI accesses your data[​](#how-ai-accesses-your-data "Direct link to How AI accesses your data")

AI tools interact with Comind.work through a structured API layer:

1. Your AI tool sends a natural language query
2. The MCP server translates it into API calls
3. Comind.work returns the data (respecting your permissions)
4. The AI tool formats and presents the results

* AI access respects your user permissions - you only see data you are authorized to view
* Responses use compact markdown/tables to minimize token usage
* Field descriptions, prefixes, and suffixes from the app schema are included so the AI understands the data context

## MCP authentication[​](#mcp-authentication "Direct link to MCP authentication")

Two authentication methods are available for connecting AI tools to your Comind.work data:

### Manual API tokens[​](#manual-api-tokens "Direct link to Manual API tokens")

The default method. Your admin provides a token that you paste into your AI tool's configuration file:

claude\_desktop\_config.json (example)

```
{
  "mcpServers": {
    "comind": {
      "url": "https://mcp.comind.work/your-org",
      "headers": {
        "Authorization": "Bearer YOUR-TOKEN-HERE"
      }
    }
  }
}
```

Manual tokens work with any MCP-compatible client. They do not expire automatically - rotate them periodically and revoke when no longer needed.

### OAuth login[​](#oauth-login "Direct link to OAuth login")

OAuth provides a browser-based sign-in flow instead of manual token management. When enabled, connecting an AI tool opens a login form where you authenticate with your Comind.work credentials.

OAuth must be explicitly enabled by a system administrator. Once enabled, all workspace members can use the login flow instead of copying tokens.

Which method to choose

Use **manual tokens** for quick setup or when OAuth is not yet enabled. Switch to **OAuth** when your team has multiple users connecting AI tools - it is easier to manage and does not require sharing tokens.

## AI tool compatibility[​](#ai-tool-compatibility "Direct link to AI tool compatibility")

| AI tool                    | MCP support   | Notes                                                                               |
| -------------------------- | ------------- | ----------------------------------------------------------------------------------- |
| Claude Desktop             | +             | Full support - recommended for Comind.work workflows                                |
| Claude Code (terminal)     | +             | Full support via `.mcp.json` config                                                 |
| VS Code (Claude extension) | +             | Full support via MCP extension settings                                             |
| ChatGPT                    | Limited       | Requires developer mode to configure custom MCP servers                             |
| Gemini                     | Not available | Google supports MCP only in developer tools, not in gemini.google.com or voice mode |

### Choosing a Claude model[​](#choosing-a-claude-model "Direct link to Choosing a Claude model")

If your AI tool lets you select a model, the choice affects both quality and cost:

* **Opus** - best for complex tasks: app development, security-sensitive changes, large context reasoning. Higher token cost, more reliable on multi-step work
* **Sonnet** - good for simple queries, quick prototyping, and data lookups. Lower cost, but may overcomplicate solutions that require precise implementation

As a rule of thumb: use Sonnet for asking questions about your data, Opus for building or modifying apps.

## Security and privacy[​](#security-and-privacy "Direct link to Security and privacy")

* **Authentication** - MCP connections require either a valid token or OAuth session tied to your user account
* **Permissions** - all queries respect workspace and app-level access controls
* **Data handling** - data is sent to your AI provider (e.g., Anthropic for Claude). Review your organization's data sharing policies before connecting AI to workspaces with sensitive data
* **Token management** - rotate manual tokens periodically and revoke when no longer needed. OAuth sessions expire automatically

MCP access can be restricted at both workspace and app levels.

## Next steps[​](#next-steps "Direct link to Next steps")

* [MCP tools](/api-integrations/ai/mcp-tools.md) - detailed setup and usage guide for MCP
* [AI Assistant](/api-integrations/ai/ai-assistant.md) - the in-app AI chat experience