Skip to main content
Builder MCP exposes PolyAI’s platform as MCP tools, grouped into the families below. Your client discovers them automatically on connection — you don’t configure tools by hand. Each tool carries its own input schema, which the client reads directly from the server. US, UK, and EU workspaces expose 97 tools; self-serve Studio workspaces expose 91 (the Alerts family is enterprise-only). Tool names appear exactly as your client sees them in tools/list.
To see exactly what your client discovered, ask it to list its available tools, or query the endpoint’s tools/list method (see Quickstart → Verify).

Build & manage agents

Create agents and manage everything they contain. Workspace- and agent-scoped, backed by the Agents API. Agents Behavior Branches Deployments Knowledge base Variants Attributes Connectors Phone numbers Config pages (real-time config)

Voice & audio

Manage the voices an agent uses and the audio it caches. Voice library Deployed voice configs Voice & disclaimer tuning Audio cache

Run, test & inspect

Exercise agents, place calls, and pull back the data they produce. Backed by the Debug Chat and Conversations APIs. Debug chat Conversations Test suites Outbound calling

Monitor deployed agents

Watch live agents and react to issues. Backed by the Alerts API. Alerts

Manage agent infrastructure

Configure the plumbing behind an agent. MCP servers on an agent Secrets

Limiting what a client can do

To narrow what a client can reach, scope the API key itself. Create a least-privilege key limited to the agents and permissions that client needs — for example, a read-only key for a monitoring assistant. The key’s scope governs which calls succeed, so a narrowly scoped key can’t create, delete, or deploy even though the tools are still discovered.
Last modified on July 13, 2026