Builder MCP

Build, test, and deploy PolyAI agents from your IDE or MCP client, without calling the API directly.

Builder MCP is PolyAI’s authenticated MCP server for building agents. Connect an MCP client — Claude Code, Cursor, Claude Desktop, or Codex — and you can create, test, and deploy PolyAI agents from your IDE without calling the REST API yourself. When the client connects, it discovers the available tools and their input schemas.

Builder MCP vs Agent Studio MCP integrations

The two point in opposite directions:

	Builder MCP	Agent Studio MCP integrations
Direction	An MCP client connects in to PolyAI	Your PolyAI agent connects out to a third-party server
Who acts as the client	Your IDE / AI coding tool	Your live PolyAI agent
What it’s for	Build, test, and deploy agents	Give a live agent extra tools during a conversation
Docs	This page	Agent Studio MCP integrations

What you can do

Builder MCP exposes PolyAI’s platform as tools grouped into three areas:

Agents

Create agents, edit behavior and knowledge, manage branches, variants, connectors, and phone numbers.

Data

Retrieve conversations and transcripts to test and inspect agent behavior.

Alerts

Create alert rules and read active alerts to monitor deployed agents.

See Capabilities for the full tool list.

Why use it

• No API plumbing. The client reads each tool and its schema from the server, so you don’t hand-write requests, hosts, or headers.
• Conversational workflows. Ask your assistant to branch an agent, update its greeting, test it, and publish to staging, and it runs the tool calls in order.
• Full deployment lifecycle. Variants, environments, and deployments are available as tools. See the walkthrough.
• One API key. A single workspace-scoped key authenticates every tool.

Get started

1. Authenticate

   Create a workspace-scoped API key.

2. Connect your client

   Add Builder MCP in one command. See Quickstart.

3. Build, test, deploy

   Follow the end-to-end walkthrough.