Skip to main content
Actions determine what your AI agent can do beyond answering questions. This is where you enable and configure the tools available to your agent.

​
What Are Actions?

Actions are the capabilities you give your agent. When a user asks something that requires taking action (like checking an order status or creating a ticket), your agent uses these tools to help.

​
Types of Actions

Crow supports three types of actions:

​
API Tools (OpenAPI)

Operations from your OpenAPI specification. These call your REST API endpoints. Example tools:
  • getOrderStatus — Look up an order
  • createTicket — Create a support ticket
  • getAccount — Retrieve account details
Set up OpenAPI integration →

​
MCP Tools

Tools from connected MCP servers. These interact with external services. Example tools:
  • Notion search and read operations
  • Custom database queries
  • Third-party service integrations
Connect MCP servers →

​
Workflows (Journeys)

Guided, multi-step processes you define. These walk users through complex tasks. Example workflows:
  • New user onboarding
  • Troubleshooting flow
  • Return request process
Create workflows →

​
Configuring Actions

​
Accessing the Actions Page

Go to Actions in the sidebar to see all available tools organized by type.

​
Enabling/Disabling Tools

Each section has toggles to control which tools your agent can use:
  1. API Tools — Check/uncheck individual endpoints
  2. MCP Servers — Toggle entire servers or individual tools
  3. Workflows — Enable/disable specific journeys

​
Select All / Deselect All

Use the quick toggle to enable or disable all tools in a section at once.

​
How the Agent Uses Actions

When a user asks a question:
  1. Agent analyzes the request
  2. Identifies relevant tools based on the question
  3. Executes the tool if needed
  4. Uses the result to formulate a response
The agent automatically decides when to use tools based on the user’s question and the tool descriptions.

​
Best Practices

​
Start Conservative

Begin with a limited set of tools and expand as needed. Too many tools can confuse the agent.

​
Use Descriptive Names

Tool names and descriptions help the agent understand when to use them. Clear descriptions lead to better tool selection.

​
Test Thoroughly

After enabling tools, test in the Sandbox with realistic user questions to verify the agent uses them correctly.

​
Review Regularly

As your product evolves, review which actions are enabled. Disable tools that are no longer relevant.

​
Tool Visibility

When tools are enabled, they appear to your agent as available capabilities. The agent sees:
  • Tool name
  • Description
  • Required parameters
  • What the tool returns
This information helps the agent decide when and how to use each tool.

​
Need Help?

Book a Call

Schedule a 30-minute call