The Skills tab of an AI Agent brings together everything that gives it additional capabilities: integrations and MCP, which let it perform actions in external systems, and logic, which controls the flow of the process. Working with external services happens in two moments: first you integrate the service into the project, meaning you link the account once, and then you connect that integration to the agent, enabling its actions and describing when to use them.

You can:

• Integrate external services into your project (Google Calendar, Gmail, Google Sheets, and others).
• Connect integrations and MCP to one or more AI Agents.
• Use your own code actions as internal MCP.
• Connect external MCP servers to reach your company’s own systems.
• Create and use logic: instruction and loop.
• Add skills and logic inside a Funnel state.

Prerequisites

Access to your Botmaker account with permissions to manage the project’s integrations.

At least one AI Agent created, the one you’ll connect these capabilities to.

The credentials for the external service account you’re going to link (for example, the Google user who manages the business calendar).

For internal MCP, code actions already created in your project. For external MCP, the MCP server’s URL and its authentication details, which the technical team that manages that server should give you.

Key concepts

Integrate: link an external service account to the project. You do this once, and it makes the integration available to all of your agents.

Connect: enable an already available integration inside an agent, selecting its actions and describing when to use them.

Action: each specific task the agent can perform in the external service (for example, check availability or create a calendar event).

MCP (Model Context Protocol): a standard that lets your AI Agent use tools from other systems without you having to program the connection from scratch. Each MCP gives the agent a set of actions it can run during the conversation. In Botmaker there are two types: internal and external.

Internal MCP: the ones that live inside Botmaker. There you’ll find the code actions you created in your project, written in JavaScript or Python.

External MCP: MCP servers that already exist outside Botmaker and that you connect by entering their URL. Their tools are already built: when you connect the server, they become available to your agent. You use them to reach your company’s own systems, for example your store’s inventory or your private CRM.

Logic: nodes that control what the agent does and how it moves forward. There are three types: instruction, conditional, and loop. They’re now managed inside Skills.

How to integrate and connect an external service

Step 1: Integrate the service into your project

Link the external service account to your project. This step authorizes Botmaker to use that service on your behalf and makes the integration available to connect to your agents afterward. You do it once per project: if you already integrated Google Calendar, you don’t need to repeat it for each agent.

Note: granting permissions to an external service means authorizing access to that account. Do it only with accounts that belong to your business, and review the permissions being requested.

Step 2: Open the agent’s Skills tab

Go into the AI Agent and open the Skills tab. There you’ll see the integrations, MCP, and logic available for that agent. An integration appearing in the list means it’s already linked to the project, not that the agent is using it — you define that in the next step.

Step 3: Connect the integration and enable its actions

Connect the integration you’re going to use and select the specific actions the agent will be able to run, for example checking availability or creating an event. Then describe when it should use each one.

That description is what lets the agent decide when to reach for each action, so it pays to be specific: instead of “check the calendar,” write “use this when the customer asks for an appointment and has already given the day they prefer.”

Note: enable only the actions the agent actually needs for its process. The fewer options it has, the more accurate its decision about which one to use at each moment.

Step 4: Test it

Test the behavior using the Test button in the orchestrator and check, in the reasoning, that the agent runs the external action when it should. The reasoning shows you which tool it picked and what data it called it with, so if the agent doesn’t use the action you expected, that’s where you’ll see why.

Note: if you’re going to connect an email inbox, the linking is done through Mailbot; then you return to AI Agents to continue.

How to connect MCP to your agent

Beyond the ready-to-use integrations, inside Skills you can give your agent your own tools through MCP. There are two ways to do it, and each solves a different need:

• Internal MCP: tools you create that live inside Botmaker, built from your code actions in JavaScript or Python.
• External MCP: tools that already exist on an MCP server outside Botmaker, which you connect by entering its URL. This is the path for your agent to reach your company’s own systems, like a store’s inventory or a private CRM.

The practical difference is where the tool lives: with internal MCP the code is yours and runs inside Botmaker; with external MCP the system is already built on your side and Botmaker simply connects to it.

Note: in both cases, adding the MCP is only half the job. You then have to connect it to the agent from Skills and describe when it should use each tool, just like any integration.

Internal MCP: use your own code actions

Go into the AI Agent, open the Skills tab, and enter Internal MCP. There you’ll see the code actions you already created in your project, in JavaScript or Python.

Each code action is a function you wrote to solve something specific: calculating a discount, validating an ID number, or building a message with your business data. By exposing them as internal MCP, your agent can run them during the conversation.

If you don’t have any code actions yet, this section will be empty. You need to create them in your project first: to learn how, we recommend the “Code actions” course in Botmaker Academy.

Connect the action you want to enable and describe when the agent should use it, with the same criteria as integrations: the more specific the description, the better the agent decides.

External MCP: connect one of your company’s MCP servers

In External MCP you connect MCP servers that already exist outside Botmaker. You don’t create the tools here: they’re already built on that server and, once you connect it, they become available to your agent.

Step 1: Click Create

Inside Skills → External MCP, click Create. The connection form will open.

Step 2: Fill in the MCP server details

You’ll find four fields. Fill them in like this:

• MCP URL: the address of the server you want to connect to. This is the field that defines which system your agent reaches, so ask the technical team that manages that server for it.
• Name: how you’ll identify this MCP inside Botmaker. Use a name that says what the system is, for example “Online store inventory” or “Internal CRM,” so it’s easy to recognize when you connect it to different agents.
• Description: what this server is for and what kind of information or actions it offers. Write it with the agent in mind: the clearer it is about what the server solves, the better the agent will decide when to use it.
• Authentication type: how Botmaker identifies itself to that server in order to be allowed to use it. Choose the one that matches the server you’re connecting; if you choose a different one than that server expects, the connection is rejected.

Step 3: Connect the MCP to the agent and enable its tools

Creating the MCP makes it available in the project, but no agent is using it yet. Go back to the AI Agent’s Skills tab, connect it, and select the specific tools you want to enable, describing when it should use them.

Step 4: Test it

Test the behavior using the Test button in the orchestrator and check, in the reasoning, that the agent calls the MCP tool when it should and that the server’s response comes back complete.

How to use logic

Logic controls how the agent’s process moves forward, and it’s managed inside Skills. There are three types:

• Instruction: tells the agent to do one specific task, described in natural language.
• Conditional: opens different paths depending on whether a condition is met.
• Loop: repeats a set of steps, for example going through a list, until a condition is met.

Besides managing them in this tab, you can add a logic node inside a Funnel state, from the Instructions for the agent section. That’s the recommended path when the logic only applies to one specific moment of the process rather than to the whole agent.

Important considerations

• Integrating is not connecting: integrating links the account to the project and happens once; connecting enables that integration inside a specific agent. If the agent can’t see a tool you already integrated, the connect step is almost always what’s missing.
• Integrations depend on the external service: if that service has an outage, the agent may report that it can’t complete the action for now. That’s not a configuration error.
• External MCP depends on a system Botmaker doesn’t manage: if the server is down, changes its URL, or modifies its tools, the agent won’t be able to complete the action either. When something behaves oddly, check that server’s status first.
• Take care with permissions and data: an external MCP can give access to sensitive business information. Connect only your own servers or those of trusted providers, and use credentials with the minimum permissions needed.
• The description is not a minor detail: in both integrations and MCP, the description of when to use each action is what guides the agent’s decision. If the agent isn’t using a tool you expected it to, start by reviewing that.

To learn more about the code actions exposed as internal MCP, we recommend the “Code actions” course in Botmaker Academy.

Remember to visit our Help Center for further information.