Let a voice agent use other apps' MCP servers during a call

1. Connect the server with POST /v1/mcp-servers. Its tools are read straight away.
2. Choose which tools agents may use with enabled_tools.
3. Attach the server to an agent with mcp_server_ids.

​

Connect a popular app

• Sign in. Press the button and sign in to the app. Wixzel Voice registers itself with the app, so there is nothing to set up first.
• API key. Paste a key or token from the app. It is sent as a header.
• Your own app. The app does not let other apps register themselves. Create an OAuth app with the vendor once, give it the redirect URL below, and paste its client id and secret.
• No sign-in. Nothing to enter, or only the server’s URL.

​

Apps that use an API key

curl https://api.voice.wixzel.com/v1/mcp-servers \
  -H "Authorization: Bearer $WIXZEL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "GoHighLevel",
    "url": "https://services.leadconnectorhq.com/mcp/",
    "auth": {
      "type": "headers",
      "headers": [
        { "name": "Authorization", "value": "Bearer pit-..." },
        { "name": "locationId", "value": "ve9EPM428h8vShlRW1KT" }
      ]
    }
  }'

​

Apps that need your own OAuth app

1. Create the app in the vendor’s developer settings:
   • HubSpot: Development, then MCP Auth Apps.
   • Salesforce: an External Client App with the scopes mcp_api and refresh_token.
   • Slack: an app in your workspace at api.slack.com/apps. Slack only lets internal or directory apps use MCP, and a workspace admin may need to approve it.
   • Asana: an MCP app in the Asana developer console.
   • Google Calendar: a Web application OAuth client in a Google Cloud project with the Google Calendar API and the Calendar MCP API enabled. The Calendar MCP server is in Google’s Workspace Developer Preview Program.
2. Add this redirect URL to it: https://api.voice.wixzel.com/mcp-oauth/callback
3. Paste its client id and client secret on the app’s page in the console and press Sign in. From the API, send them as auth.oauth.client_id and auth.oauth.client_secret, then start the sign-in as below.

​

Things to know about some apps

• GoHighLevel has no tool to book an appointment yet. Use Cal.com, Calendly or the built-in booking for that.
• Calendly has no reschedule tool; the agent cancels and books again. Booking a slot needs a paid Calendly plan.
• Close is signed in with read and safe-write access. Updating existing records needs Close’s destructive scope, which is not asked for.
• Help Scout is read-only. It cannot create tickets or notes.
• Stripe: live or test mode is chosen on Stripe’s sign-in screen. Refunds wait for someone to approve them in Stripe.
• Razorpay cannot make refunds over its hosted server.
• Dodo Payments asks for your API key and test or live mode when you sign in. Its execute tool can do anything the key can, including refunds, so it starts switched off.
• Zapier charges 2 Zapier tasks for each tool call. Choose the actions in Zapier, then refresh the tools.
• Make scenarios should answer within 25 seconds to be useful on a call.
• Freshdesk limits MCP actions by plan.
• Home Assistant must be reachable at a public https address. Home Assistant Cloud gives every instance one; homeassistant.local cannot be reached. Add the Model Context Protocol Server integration in Home Assistant first. The agent only sees the devices exposed to voice assistants. See Smart home devices.
• Exa uses Exa’s free, rate-limited tier. No account is needed.

​

Connect any MCP server

curl https://api.voice.wixzel.com/v1/mcp-servers \
  -H "Authorization: Bearer $WIXZEL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Orders",
    "url": "https://mcp.example.com/mcp",
    "auth": {
      "type": "headers",
      "headers": [{ "name": "Authorization", "value": "Bearer sk_live_..." }]
    }
  }'

const server = await wixzel.mcpServers.create({
  name: 'Orders',
  url: 'https://mcp.example.com/mcp',
  auth: { type: 'headers', headers: [{ name: 'Authorization', value: 'Bearer sk_live_...' }] },
});
console.log(server.status, server.tools.map((t) => t.name));

​

Authentication

• none for a public server.
• headers for an API key or any fixed header. Values are encrypted at rest and never returned; responses list only the header names. On update, send a header without value to keep the stored one.
• oauth for a server that signs in. Most hosted MCP servers do.

​

Servers that sign in with OAuth

curl -X POST https://api.voice.wixzel.com/v1/mcp-servers/$ID/oauth/start \
  -H "Authorization: Bearer $WIXZEL_API_KEY"

{
  "object": "mcp_server_oauth",
  "authorization_url": "https://auth.example.com/authorize?client_id=...&state=...",
  "redirect_uri": "https://api.voice.wixzel.com/mcp-oauth/callback",
  "expires_at": "2026-10-02T10:10:00.000Z"
}

​

Choose the tools

curl -X PATCH https://api.voice.wixzel.com/v1/mcp-servers/$ID \
  -H "Authorization: Bearer $WIXZEL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "enabled_tools": ["lookup_order", "create_note"] }'

​

Attach it to an agent

curl -X PATCH https://api.voice.wixzel.com/v1/agents/$AGENT_ID \
  -H "Authorization: Bearer $WIXZEL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "mcp_server_ids": ["'$ID'"] }'

​

What happens on a call

• The tools are declared to the model when the call starts, from the stored list, so connecting a server never delays the greeting. The connections open in the background.
• Each tool is named <app>__<tool>, for example orders__lookup_order, and its description starts with the app’s name.
• The agent says a few words before it uses a tool, so the caller is not left in silence. On the classic and Sarvam engines, if the model goes straight to the tool, the agent says “One moment, let me check that.” for it.
• Before using a tool the server marks as changing data, the agent asks the caller to confirm.
• A tool call times out after 10 seconds. Failures become a sentence the agent can say (“I couldn’t reach Orders right now”); they never end the call.
• Results are cut to 4,000 characters before the model sees them.
• At most 20 tool calls per call, and three rounds of tool calls per caller turn before the agent has to answer in words.

​

Cost

​

Security

• Allowlist. Only enabled tools are declared, and a call to any other name is refused even if the model guesses it.
• Untrusted results. The agent is told that tool results are data, not instructions.
• Network. Server URLs must be public https (or http) on port 443 or 80. Private, loopback and link-local addresses are refused when the server is saved and again on every request, including each redirect, and the connection goes to the address that was checked. A redirect to another host does not carry your headers.
• Secrets. Header values, OAuth tokens and client secrets are encrypted at rest and never returned by the API.
• Scopes. Managing servers needs mcp_servers:read and mcp_servers:write.

​

Smart home devices

• Do not expose locks, alarms or garage doors to voice assistants in Home Assistant.
• Only the read tool (GetLiveContext) is switched on when it is connected. Switch on the tools that change devices only for an agent whose prompt checks who it is talking to.