Tools in Creatio AI Studio
A tool is a single capability an agent can call. Tools are the only building block that lets an agent act on the outside world — a knowledge source informs an answer but never changes anything. Learn more: Choose between a skill, a tool, and a knowledge source.
Tools are enabled per agent, on two tabs of the Build tab group. The tools the platform and the model provide are on the Tools tab. The tools that come from an external system, over HTTP or MCP, are on the Integrations tab, where you enable them integration by integration.
Tool types
Type | Description |
|---|---|
Built-in model tools | Capabilities the model itself provides. One checkbox, no setup, on the Tools tab. Availability depends on the model you selected. |
HTTP | A REST endpoint you describe yourself — base URL, path, method, and how to read the response. Enabled on the Integrations tab. Learn more: Add an HTTP tool. |
MCP | A tool discovered from an MCP server through an integration. The server advertises what it offers and you enable the ones the agent may use, on the Integrations tab. Learn more: MCP integration. |
Generative UI | A tool that returns an interactive card rendered in the conversation instead of plain text. |
Built-in model tools
Tool | Description |
|---|---|
Web search | Lets the model search the web and answer with cited sources. Works across OpenAI, xAI, Anthropic, and Gemini models. |
File search | Lets the model search uploaded files and vector stores. Availability depends on the selected model. |
Image generation | Lets the model generate images. Availability depends on the selected model. |
Code execution | Lets the model run code in a sandbox to compute an answer. Availability depends on the selected model. |
X / Twitter search | Lets the model search X. Available on xAI Grok models only. |
Risk levels
Every tool carries a risk level that describes what it can do.
Risk level | Meaning |
|---|---|
Read | The tool only fetches data. |
Write | The tool changes something. |
Admin | The tool performs an administrative operation. |
The risk level describes the tool; it does not gate it. To make an agent ask before it runs a tool, create a tool confirmation policy, which is where the approval and its routing are set. Learn more: Set up tool confirmation.
The Tools tab
The Tools tab lists the platform and model tools available to the agent, in the groups Built-in model tools, Generative UI tools, Activity tools, Built-in tools, and Other tools. Enable one by selecting its checkbox. A new agent starts only with the built-in tools its model supports, such as Web search, so an agent you have not edited can already reach the open web. Clear Web search if the agent should answer only from what you gave it.
The chips above the list, such as "Available," "Enabled," "HTTP," "MCP," and "Generative UI," report how many tools are in each state. They are counters, and clicking one does nothing.
The Integrations tab
The Integrations tab lists the tools that external systems expose, grouped by the integration they come from. Each integration starts with none of its tools enabled, so a newly connected system reads "0 of 10 enabled" until you choose. Select an Access mode for the integration, then select the Enabled checkbox on each tool the agent may call.
Filter the list by All, Enabled, Available, or Disabled by admin. A tool can report "Not available in this environment" or "Disabled by admin," in which case it cannot be enabled for the agent.
An integration can be restricted on its Access tab. A restricted integration is hidden from users who have no grant for it, and they cannot enable its tools. Runs that start without a signed-in user, such as channel or triggered runs, can use only integrations that are open to the whole organization. Learn more: Integrations.