Before you start: what MCP gives your agent
MCP (Model Context Protocol) lets you give a MyChatBot agent custom tools, your own internal API, a pricing engine, an inventory service, anything. Before connecting, have three things ready: the MCP server endpoint, its auth credentials, and a clear idea of which actions the agent should be allowed to call. A tightly-scoped tool beats a powerful one nobody bounded.
Step 1, Connect the MCP server
In MyChatBot, open the agent's tools and add an MCP server. Paste the endpoint and authorize, auth is where the agent proves it is allowed to talk to your server. Connection takes seconds, and the server is now reachable by the agent.
Step 2, Capability discovery
Once connected, MyChatBot performs capability discovery: it reads the tools your MCP server exposes and makes them available to the agent. You see the full list of actions, query inventory, create an order, fetch a quote, without writing any glue code. The agent now knows what it can do.
Step 3, Permission scoping
This is the most important step. Decide exactly which discovered tools the agent may call, and under what conditions, read-only lookups freely, write actions only after confirmation, sensitive ones behind Hand-off Control. Permission scoping is how you give an agent power without giving it free rein.
Step 4, Graceful degradation and go live
Set what happens if the MCP server is slow or down: the agent should degrade gracefully, fall back to its Knowledge Base, offer a callback, or hand off to a human via Flight Control, never stall or invent an answer. Let the Configuration Wizard battle-test the tool calls, turn it on, and watch the first real interactions. Five minutes to connect; the payoff is an agent that can actually act in your systems.