***
title: Use your MCP request to interact with an MCP server
updated: 2026-01-12T00:00:00.000Z
max-toc-depth: 2
----------------
You can interact with [Model Context Protocol](https://modelcontextprotocol.io/introduction) (MCP) servers with server and client capabilities. These features give AI models further context and extend their capabilities. Servers that support the [MCP Apps Extension](https://github.com/modelcontextprotocol/ext-apps) can also return interactive user interfaces.
*Server capabilities* are features the server exposes. These capabilities include [tools](https://modelcontextprotocol.io/docs/concepts/tools), [resources](https://modelcontextprotocol.io/docs/concepts/resources), and [prompts](https://modelcontextprotocol.io/docs/concepts/prompts). *Client capabilities* are features the client exposes. They typically require user input. These capabilities include [sampling](https://modelcontextprotocol.io/docs/concepts/sampling), [roots](https://modelcontextprotocol.io/docs/concepts/roots), and [elicitation](https://modelcontextprotocol.io/docs/concepts/elicitation).
When you send requests to a server, the server returns a response. Servers can also send real-time notifications and requests of their own. You can use these interactions to experiment, test, and evaluate a server.
## Interact with an MCP server capability
Server capabilities include the following:
{/* vale Microsoft.FirstPerson = NO */}
{/* First-person use is okay (part of an example prompt). */}
* *Tools* are actions the server can perform for an AI model. For example, an AI model can use a server's tool to create a calendar event for you.
* *Resources* are data the server can retrieve for an AI model. For example, an AI model can use a server's resources to get your schedule from your calendar.
* *Prompts* are prompt templates the server can suggest for interacting with an AI model. For example, the server can suggest, "Can you create an event for me?" or "Can you show me my schedule for tomorrow?"
{/* vale Microsoft.FirstPerson = YES */}
To interact with an MCP server capability, do the following:
1. Choose an existing MCP request or [create a new one](/docs/postman-ai/mcp-requests/create/).
2. Click **Load Capabilities**. If prompted, review the consent prompt and grant Postman access. Postman connects to the server and displays the server's capabilities.
3. List the server's capabilities:
* To list the [tools](https://modelcontextprotocol.io/docs/concepts/tools) capabilities, open the **Tools** tab.
* To list the [resources](https://modelcontextprotocol.io/docs/concepts/resources) capabilities, open the **Resources** tab.
* To list the [prompts](https://modelcontextprotocol.io/docs/concepts/prompts) capabilities, open the **Prompts** tab.
You can also 
**Search tools, resources, and prompts**.
4. Choose a capability and define its arguments. Postman creates a JSON message for you.
5. Click **Run**. Postman sends your request.
Postman displays the server's response in the **Response** tab. Postman also displays real-time notifications in the **Notifications** tab. You can search, filter, and clear messages. Each message includes a type and timestamp. To see its details, click the message.
You can also see whether you're connected or disconnected. To disconnect from the server, click **Disconnect** next to **Run**. To reconnect, click **Connect**.
## Interact with an elicitation
Client capabilities include elicitation. *Elicitation* is when the server requests information from you. For example, if you prompt an AI model to create a calendar event for you, the server can request missing information, such as the date and time.
To interact with an elicitation, do the following:
1. [Interact with an MCP server capability](#interact-with-an-mcp-server-capability) that triggers an [elicitation](https://modelcontextprotocol.io/docs/concepts/elicitation) request.
2. In the response pane, complete the **Elicitation** form. Postman creates a JSON response for you.
3. Click **Accept**. Postman sends your response. You can also **Deny** or **Cancel** the request.
## Interact with sampling
Client capabilities include sampling. *Sampling* is when the server prompts an AI model to complete a task it can't complete itself. For example, if a server can access your calendar, but can't summarize your weekly schedule, it can prompt an AI model to summarize it for you.
With Postman, you mock the AI model's response to a sampling request.
To interact with sampling, do the following:
1. [Interact with an MCP server capability](#interact-with-an-mcp-server-capability) that triggers a [sampling](https://modelcontextprotocol.io/docs/concepts/sampling) request.
2. In the response pane, complete the **Sampling** form. You're mocking the AI model's response. Postman creates a JSON response for you.
3. Click **Accept**. Postman sends your response. You can also **Deny** the request.
{/*
## Interact with roots
Client capabilities include roots. *Roots* is when the server requests access to local files or directories. For example, if you prompt an AI model to download an event for you, the server can request access to your downloads folder.
To interact with roots, do the following:
1. [Interact with an MCP server capability](#interact-with-an-mcp-server-capability) that triggers a [roots](https://modelcontextprotocol.io/docs/concepts/roots) request.
1. In the response pane, complete the **Roots Form**. Postman creates a JSON response for you.
1. Click **Accept**. Postman sends your response. You can also **Deny** or **Cancel** the request.
*/}
## Interact with an MCP App
MCP Apps is an extension of the Model Context Protocol. It standardizes how servers deliver interactive user interfaces. For example, you can view data visualizations, such as interactive charts and graphs, directly in the response.
To interact with an MCP App, do the following:
1. [Interact with an MCP server capability](#interact-with-an-mcp-server-capability) that supports the [MCP Apps Extension](https://github.com/modelcontextprotocol/ext-apps).
2. If preview's not already open, in the response pane, click 
**Preview**.
3. Explore the server's interactive user interface.