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

# MCP Servers Reference

> Endpoints to connect MCP servers to agents — expand AI action capabilities with external tools

## Concept

**MCP Servers** (Model Context Protocol) are external servers that expose tools and resources so that Timely.ai agents can perform real-world actions during a conversation. Instead of the agent simply responding with text, it can use MCP tools to query a database, create a calendar event, look up an order in e-commerce, send an email, or any other action the MCP server supports.

The MCP protocol is an open standard — any compatible server can be connected to a Timely agent. You can use publicly available community servers or develop your own to integrate your company's internal systems.

## How the Integration Works

```
customer sends message
    │
    ▼
AI agent processes and decides to use a tool
    │
    ▼
Timely.ai calls the MCP Server with the tool parameters
    │
    ▼
MCP Server executes the action and returns the result
    │
    ▼
agent incorporates the result into the response to the customer
```

## MCP Server Lifecycle

```
add server (URL + configuration)
    │
    ▼
connect → Timely handshakes with the server
    │
    ▼
list available tools
    │
    ▼
activate specific tools on the agent
    │
    ├── sync tools (update list)
    │
    ├── deactivate specific tools
    │
    └── disconnect / delete
```

## Main Fields

| Field          | Type          | Description                                    |
| -------------- | ------------- | ---------------------------------------------- |
| `id`           | string (UUID) | Unique MCP Server ID in the workspace          |
| `agent_id`     | string (UUID) | Agent it is connected to                       |
| `name`         | string        | Descriptive server name                        |
| `url`          | string        | MCP endpoint URL (SSE or HTTP)                 |
| `transport`    | enum          | `sse` \| `http`                                |
| `status`       | enum          | `connected` \| `disconnected` \| `error`       |
| `auth_type`    | enum          | `none` \| `bearer` \| `api_key`                |
| `auth_token`   | string        | Authentication token (write-only)              |
| `tools_count`  | number        | Number of tools available on the server        |
| `active_tools` | array         | List of slugs of tools activated on the agent  |
| `created_at`   | string        | ISO 8601 timestamp when added to the workspace |

## Tool Fields

| Field         | Type    | Description                                                   |
| ------------- | ------- | ------------------------------------------------------------- |
| `slug`        | string  | Unique tool identifier on the server                          |
| `name`        | string  | Human-readable tool name                                      |
| `description` | string  | What the tool does (used by the LLM to decide when to use it) |
| `parameters`  | object  | JSON Schema of the tool parameters                            |
| `active`      | boolean | Whether the tool is activated for the agent                   |

## Available Endpoints

<Note>
  See the left sidebar for the full list of endpoints in this group. Each endpoint has its own page with request/response schema, cURL/JS/Python examples, and an interactive playground.
</Note>

## Next Steps

<Card title="Your first request" href="/en/start/quickstart-api">
  If you haven't made your first call yet, start here.
</Card>