Chatley AI — AI Agents for Calls, Chat & Messaging
Chatley MCP Server for AI Voice Agent Management

Chatley MCP Server for AI Voice Agent Management

By Chatley AI · October 5, 2026 · Product · Connect Claude, ChatGPT, or Cursor to Chatley with the MCP server. Manage AI voice agents, calls, knowledge bases, and campaigns from your AI client.

mcp serverai voice agentschatley aiai agent managementvoice ai automation

The Chatley AI MCP Server lets any MCP-compatible AI client read your calls, transcripts, and analytics, and draft changes to your AI voice agents. Chatley AI is an AI voice receptionist and white-label agent platform, with one AI agent across voice, chat, and SMS. The MCP server is how developers connect to it. You sign in once in a browser. Every agent change saves as a draft you review before it goes live.

How do I connect my AI client to Chatley AI?

Add https://mcp.chatley.ai/mcp as a remote MCP server in your AI client. The server uses Streamable HTTP and falls back to SSE for older clients. HTTPS is required, so plain HTTP requests are rejected.

After you add the URL, your client walks you through signing in to Chatley AI in a browser window. There is no API key to copy, no manual registration step, and nothing to paste back into the client. You sign in once, and your assistant can start reading call data right away.

The server works with any MCP-compatible AI client. Some clients have their own setup details, and the Connecting a client page in the Chatley AI docs lists them. If your assistant already handles your calendar, your CRM and your inbox, connecting Chatley AI adds your voice agents to the same conversation, so one session can pull a transcript and send a summary to a teammate.

What can my assistant do through the MCP server?

The server exposes 16 tools, grouped by the area they cover. The Tool reference lists every parameter and what each tool returns.

Calls

List calls, read one call’s details, read its transcript, and pull aggregate analytics

AI voice agents

List agents, read one agent’s full configuration, create an agent, save changes as a draft, and publish that draft

Voices

List the available voices

Knowledge

List knowledge bases and delete them

Account

List your workspaces and the Chatley AI accounts you administer

Metered access

Read the rate card and your balance

A typical session starts with a plain request. Ask your assistant to list last week’s calls, then have it read the transcripts of the ones that ended without a resolution and summarize what callers asked for. Next, ask it to read the configuration of the agent that handled those calls and draft a fix to its instructions. The assistant saves the fix as a draft, and you decide when to publish it. We keep the tool list short so your assistant picks the right tool for each request.

How does the server stop an AI agent from breaking things?

An AI agent working at speed carries a different risk profile from a person clicking through a dashboard. The server leaves out or limits the operations with the largest potential damage, and 3 rules apply.

Nothing is deleted in a single step. Removing data always takes a second, explicit confirmation, so one misread instruction cannot erase a knowledge base.

Live AI voice agents are never edited directly. Your assistant saves changes as a draft, and going live is a separate action you can review first.

There are no bulk-delete or update-everything-matching tools. Those are the highest-risk operations to put in an agent’s hands, so the server does not offer them. Together, these rules let you give an assistant broad access to your account while a review step sits in front of every change to a live agent. The Safety and limits page in the Chatley AI docs explains each rule and how to work with it.

What can’t the MCP server do yet?

Contacts, campaigns and adding items to a knowledge base are not available through the server yet. If you ask your assistant for one of those, it tells you the tool does not exist and does not substitute a different action. Until a tool ships, handle those tasks in Chatley AI directly.

The Chatley AI team adds tools based on real use, so the list stays short and each tool does one job well. That approach also explains why bulk-delete tools stay out even if you request them. A smaller set keeps your assistant’s choices accurate and its responses fast.

You can track additions on the What’s new page in the Chatley AI docs. If a missing tool blocks your workflow, check that page before you build around the gap, since the tool may have shipped since you last connected. The full Chatley AI MCP server documentation covers every area in more detail.

FAQ

Frequently asked questions

No. You sign in to Chatley AI in a browser window when you add the server, and there is nothing to copy or paste back.

Ready to grow your business with AI?

Talk to our team and find the right fit for your calls, leads, and customer experience.

Start Free Trial