Diga MCP server for AI assistants

Use Diga’s hosted Model Context Protocol (MCP) server to build and manage voice agents from compatible AI assistants and development tools.

This page explains how to connect an assistant to Diga. To give a Diga voice agent access to an external MCP server during a call, see MCP integrations.

Prerequisites

Before you begin, you need:

A Diga account with administrator access to a project
A Diga API key from that project
An MCP client that supports remote servers over Streamable HTTP

If you do not have a key yet, follow Create an API key. Create a restricted key with only the permissions your assistant needs.

What you can do

The Diga MCP server turns Diga API operations into tools your assistant can discover and call. Depending on your API key permissions, you can use natural language to:

Create, update, publish, list, and inspect voice agents and their versions
Create calls and retrieve call details, messages, recordings, logs, and traces
Create and maintain knowledge bases from text, URLs, and files
Manage phone numbers, voices, integrations, webhooks, and workflows
Review project members, usage, billing, and other project resources

Your MCP client shows the exact tools available to it, including each tool’s description and required parameters.

How authentication and project selection work

Your MCP client connects to Diga over Streamable HTTP and sends your API key with every request.

Setting	Value
Server name	`diga`
Server URL	`https://mcp.diga.io/mcp`
Authentication header	`x-api-key: <DIGA_API_KEY>`

An API key belongs to the Diga project where you created it. That project is the company the assistant manages. To work with another company, create or copy a key from that project and replace the key in your MCP client configuration.

Connect your MCP client

Choose your client and replace <DIGA_API_KEY> with your complete Diga API key.

Cursor
Claude Desktop
Claude Code
Codex
Other clients

Open Cursor Settings → MCP → Add new global MCP server, then add:

{
  "mcpServers": {
    "diga": {
      "url": "https://mcp.diga.io/mcp",
      "headers": {
        "x-api-key": "<DIGA_API_KEY>"
      }
    }
  }
}

Open Settings → Developer → Edit Config, then add:

{
  "mcpServers": {
    "diga": {
      "url": "https://mcp.diga.io/mcp",
      "headers": {
        "x-api-key": "<DIGA_API_KEY>"
      }
    }
  }
}

Restart Claude Desktop after saving the file. Availability of remote server headers can vary by Claude Desktop version.

Run:

claude mcp add --transport http diga \
  --header "x-api-key: <DIGA_API_KEY>" \
  https://mcp.diga.io/mcp

Add the server to ~/.codex/config.toml:

[mcp_servers.diga]
url = "https://mcp.diga.io/mcp"
bearer_token_env_var = "DIGA_API_KEY"

Then set the environment variable before starting Codex:

export DIGA_API_KEY="<DIGA_API_KEY>"

Codex sends this value as Authorization: Bearer <DIGA_API_KEY>, which the Diga MCP server also accepts.

Configure a remote Streamable HTTP server with these values:

{
  "name": "diga",
  "url": "https://mcp.diga.io/mcp",
  "headers": {
    "x-api-key": "<DIGA_API_KEY>"
  }
}

Never put an API key in the server URL or commit it to source control. URLs can appear in browser history, proxy logs, and analytics. Prefer environment-variable support when your MCP client provides it.

Verify the connection

After saving the configuration, restart or reload your MCP client. Check that a server named diga appears and that its tools are available. Start with a read-only request such as:

List my Diga agents.

Your client should ask for approval before running a tool if its permission settings require confirmation.

Troubleshooting

The server returns 401 Unauthorized

Confirm that you copied the complete API key and that the client sends it as x-api-key without the Bearer prefix. If you use the Authorization header instead, its value must be Bearer <DIGA_API_KEY>.

A tool returns 403 Forbidden

The API key is valid but does not have permission to use that resource. Update the key’s restrictions in Settings → Project → API Keys, or use a different restricted key with the required permission.

The assistant shows the wrong company or no resources

The API key determines the project. Replace it with a key created in the intended Diga project, then restart the MCP client so it reloads the configuration.

The client cannot connect to the server

Verify that the URL is exactly https://mcp.diga.io/mcp and that the client supports remote MCP servers using Streamable HTTP. Some clients require a restart after configuration changes.

Security recommendations

Create a separate API key for each MCP client and environment
Grant only the read or write permissions the assistant needs
Review tool calls before approving actions that change or delete resources
Rotate the key immediately if it is exposed

​Prerequisites

​What you can do

​How authentication and project selection work

​Connect your MCP client

​Verify the connection

​Troubleshooting

​Security recommendations

Prerequisites

What you can do

How authentication and project selection work

Connect your MCP client

Verify the connection

Troubleshooting

Security recommendations