Introduction

Manage MCP tools via Ada's MCP Server

From an AI assistant connected to Ada’s MCP Server (Claude, ChatGPT, or any MCP-compatible client), you can see your MCP tool servers, list the tools on each, and change which tools the AI Agent may use, without opening the dashboard.

Connect an assistant first. See Getting started. For what MCP tools are and how they work, see the MCP tools overview.

What you can do

  • See your servers and tools. Ask the assistant to list your connected MCP servers and the tools on each, with each tool’s enabled status. This is read-only and takes effect immediately. See list_entities.
  • Enable or disable a tool. Change which tools the Agent may call.
  • Set how a tool authenticates. Choose whether a tool runs on the connected account, or requires the end user to sign in.

How changes go live

Enabling, disabling, or changing how a tool authenticates moves through a change set. The assistant stages the change, and nothing affects the live Agent until the change set is promoted. This is the same staged flow used for other agent-behavior changes, so you can review before anything goes live. Reading your servers and tools needs no change set.

What stays in the dashboard

Adding a new server is done in the dashboard, not from the assistant:

  • Connecting a server by URL, and setting its authorization (OAuth or API key).
  • The OAuth sign-in a new connection needs, which cannot run as a tool call.

Once a server is connected, which tools are on and how each authenticates can be managed from either the assistant or the dashboard.