Use the MCP server with Claude Code

Claude Code, Anthropic's command-line coding agent, can connect to the MCP server so that it sends test messages, lists WhatsApp templates and reads delivery reports on your account while it helps you build an integration. This page shows the setup. Claude Code's own MCP documentation is the reference for its commands and file formats: Connect Claude Code to tools via MCP.

Before you start#

1. Get a key: in the admin portal open Integrations → MCP Integration and choose Get temporary key. Use a temporary key while you experiment.
2. Put it in an environment variable rather than typing it into commands you might share:

export BROADNET_TOKEN="paste-your-key-here"

Add the server for yourself#

claude mcp add --transport http omnichannels https://mcp.broadnet.me/mcp \
  --header "Authorization: Bearer $BROADNET_TOKEN"

Your shell expands $BROADNET_TOKEN when you run the command, so the key is stored in your local Claude Code configuration (~/.claude.json). Without a scope flag the server is available only to you in the current project. Add --scope user to make it available in every project.

Check the connection:

claude mcp list
claude mcp get omnichannels

Inside a Claude Code session, the /mcp command shows the server's status and its tools. To remove the server, run claude mcp remove omnichannels.

Share it with your team#

To give everyone working in a repository the same server without sharing a key, commit a project-scoped .mcp.json file and let each developer supply their own token through an environment variable. Claude Code expands ${VAR} in the url and headers fields.

{
  "mcpServers": {
    "omnichannels": {
      "type": "http",
      "url": "https://mcp.broadnet.me/mcp",
      "headers": {
        "Authorization": "Bearer ${BROADNET_TOKEN}"
      }
    }
  }
}

You can also create this file with claude mcp add --transport http omnichannels --scope project https://mcp.broadnet.me/mcp --header 'Authorization: Bearer ${BROADNET_TOKEN}'; the single quotes keep your shell from expanding the variable. Claude Code asks each developer to approve project-scoped servers before it uses them.

Control which tools run#

By default Claude Code asks before it uses an MCP tool. Keep it that way for anything that sends: a send tool sends a real, billed message to a real person.

Permission rules in Claude Code's settings refer to MCP tools as mcp__<server>__<tool>. With the server named omnichannels, for example:

{
  "permissions": {
    "allow": ["mcp__omnichannels__get_*"],
    "deny": []
  }
}

This lets tools whose names start with get_ (read tools such as the delivery report lookup) run without a prompt, while every other tool, including all send tools, still asks. Tool names depend on the server version and your account, so check them under /mcp before you write rules. Add a tool to deny to remove it completely. See Claude Code's permissions documentation for the full rule syntax.

Things to ask#

• "Use the omnichannels tools to list my approved WhatsApp templates, then write a TypeScript function that sends the order_update template with the fields it needs."
• "Send a test SMS from OmniDemo to my number 447700900123, then check its delivery report."
• "Read the delivery report for yesterday's campaign and summarise the failure reasons."

For writing integration code, also give Claude Code the OpenAPI spec so it uses the exact request fields.

Troubleshooting#

To check the endpoint and key outside Claude Code, run the initialize request shown on the MCP server page.

Next steps#

• MCP server: endpoint, keys and safe-use rules for every client.
• Use the MCP server with Cursor.
• OpenAPI 3.1 spec: the REST API behind the tools.