> For the complete documentation index, see [llms.txt](https://epsilla-inc.gitbook.io/epsilladb/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://epsilla-inc.gitbook.io/epsilladb/application/mcp.md).

# MCP

**Model Context Protocol (MCP)** is an emerging open standard that enables interoperability between AI agents and external tools through tool-based API interfaces. MCP allows agents to:

* Discover external tools from a registered endpoint.
* Invoke tools (APIs) dynamically based on user intent.
* Register themselves as callable tool servers (like plugins) for other agents/platforms.

<figure><img src="/files/be4d1zTm0bEownVT2lbp" alt=""><figcaption></figcaption></figure>

Epsilla fully supports both **consuming** external MCP tool servers and **publishing** your own agent as a MCP-compatible tool server.

### Connecting Epsilla Agents to External MCP Servers

To invoke external tools via MCP inside an Epsilla agent workflow, follow these steps:

**1. Use Completion Using Tools**

From the **LLM** section, add the **Completion Using Tools** component.

<figure><img src="/files/HcPEoSkBb0KlLmpC7uDA" alt=""><figcaption></figcaption></figure>

This component is the core of an AI agent that takes user input, reasons and plans, and takes actions by using available tools.

<figure><img src="/files/spAv5RfX9Oj33LGXYoSb" alt=""><figcaption></figcaption></figure>

**2. Add MCP Server Component**

From the **“Add Components”** panel, expand **Tools**, and select **MCP Server**.

<figure><img src="/files/rHgWyxc6rJUJOlCfm76C" alt=""><figcaption></figcaption></figure>

<figure><img src="/files/5W4097hZAkiBE5wRQf6d" alt=""><figcaption></figcaption></figure>

**3. Configure MCP Server Connection**

Provide the following fields:

* **Server Name**: Friendly name (e.g., *Google Drive Connector* or *CES 2025*).
* **Server URL**: The MCP endpoint URL (e.g., `https://mcp.pipedream.net/...`).
* **Authorization Token** (if required): Token needed to access private MCP servers.

<figure><img src="/files/1hkf7QXMb2iqTMw2dB7E" alt=""><figcaption></figcaption></figure>

**4. Wire the Workflow**

Connect the MCP server's Tools output to the LLM component's Tools input.

<figure><img src="/files/94mmFU5g4r6JQV2kJuv9" alt=""><figcaption></figcaption></figure>

The agent will now dynamically select and invoke appropriate external tools via MCP based on user queries.

<figure><img src="/files/AqLUI44cmdwZH0nFfYtd" alt=""><figcaption></figcaption></figure>

<figure><img src="/files/8dutNnpHtOKE30n1OUwk" alt=""><figcaption></figcaption></figure>

### Publishing Your Epsilla Agent as an MCP Server

Epsilla allows you to **expose any agent** as an MCP-compatible tool server so it can be used by other platforms (like Cursor, Claude Desktop, **another Epsilla AI agent**, and other MCP clients).

Ensure that the access control setting is configured as "Anyone with the link."

<figure><img src="/files/PmAB6a6sIbi3G4owh4iU" alt=""><figcaption></figcaption></figure>

Once you’ve published your AI agent, you’ll see the **MCP server config** available:

<figure><img src="/files/fn4PI87tvHIAhxbC50Ic" alt=""><figcaption></figcaption></figure>

Copy the config snippet and paste it into your target MCP-supported platform (such as Cursor or other compatible apps).

Test the connection to confirm your agent is properly registered and discoverable.

<figure><img src="/files/LoQhD3rDRlTDT1taVoWk" alt=""><figcaption></figcaption></figure>

Now, your agent can be called by the other platforms as an extension to their capability.

<figure><img src="/files/SJRMYznSP4lDyZ1SHEeM" alt=""><figcaption></figcaption></figure>

This integration expands your agent’s reach—letting it operate as part of a **broader AI ecosystem** with minimal setup.