Skip to main content
Once you’ve built and published a tool, there are several ways to execute it. Choose the method that best fits your use case. Launching a tool

​
Launch tab

The most direct way to launch a tool is through the Launch tab in the Cargo interface.

​
Single run

Execute a tool once with specific parameters:
  1. Navigate to your tool in the Cargo workspace
  2. Click on the Launch tab
  3. Select Single mode
  4. Enter the required input parameters
  5. Click Run tool to execute immediately
Single runs are perfect for testing tools or processing one-off requests without setting up a full workflow.

​
Bulk run

Process multiple records at once:
  1. Navigate to your tool and select the Launch tab
  2. Choose Bulk mode
  3. Upload a CSV file containing multiple input records
  4. Map the CSV columns to your tool’s input fields
  5. Click Continue to process the entire batch
Your CSV should have headers that match your tool’s input field names. For example: 
company_domain,contact_email
acme.com,[email protected]
globex.io,[email protected]

​
Through a play

Integrate tools into automated workflows by adding them as actions in a play.
  1. Create or edit a play
  2. Add your tool as a node in the play
  3. Map inputs from previous play nodes or enrolled data
  4. Configure what happens with the tool’s output
The tool executes automatically when the play runs, and its results can be passed to subsequent actions. This is ideal for building multi-step automation sequences.

Learn more about plays

See how to build plays that orchestrate multiple tools together.

​
Through an agent

AI agents can autonomously decide when and how to use your tools.
  1. Open your agent’s configuration
  2. Add your tool to the agent’s available tools
  3. The agent will automatically launch the tool when relevant to the conversation
Tool results feed into the agent’s reasoning process, enabling sophisticated workflows where the agent chains multiple tools together based on context.

Configure agent tools

Learn how to give agents access to your tools.

​
Through an MCP server

Model Context Protocol (MCP) allows external AI systems to discover and use your tools.
  1. Enable MCP server in your Cargo workspace settings
  2. Configure which tools are exposed via MCP
  3. Connect your MCP client (Claude Desktop, VS Code, etc.)
  4. Your tools appear as available actions in the external system
This is powerful for integrating Cargo tools with AI assistants and development environments.

​
Through API calls

Execute tools programmatically via the REST API for maximum flexibility and integration options.

​
Getting your API credentials

  1. Navigate to Settings → API in your Cargo workspace
  2. Generate or copy your API token
  3. Find your tool’s ID in the tool editor URL or settings

​
API endpoint

POST https://api.getcargo.io/v1/tools/{tool_id}/execute

​
Example request

curl -X POST "https://api.getcargo.io/v1/tools/{tool_id}/execute?token={your_token}" \
  -H "Content-Type: application/json" \
  -d '{
    "company_domain": "acme.com",
    "contact_email": "[email protected]"
  }'

​
Example response

{
  "execution_id": "exec_abc123",
  "status": "completed",
  "output": {
    "company_name": "Acme Corporation",
    "employee_count": 500,
    "industry": "Technology"
  }
}

​
Async execution

For long-running tools, use async mode: 
curl -X POST "https://api.getcargo.io/v1/tools/{tool_id}/execute?token={your_token}&async=true" \
  -H "Content-Type: application/json" \
  -d '{"company_domain": "acme.com"}'
This returns immediately with an execution_id that you can poll for results: 
curl "https://api.getcargo.io/v1/executions/{execution_id}?token={your_token}"

​
Webhook callbacks

Configure a webhook URL to receive results when the tool completes: 
curl -X POST "https://api.getcargo.io/v1/tools/{tool_id}/execute?token={your_token}" \
  -H "Content-Type: application/json" \
  -d '{
    "company_domain": "acme.com",
    "webhook_url": "https://your-server.com/cargo-callback"
  }'

​
Choosing the right launch method

MethodBest forAutomation level
Launch tabTesting, ad-hoc execution, bulk importsManual
PlaysScheduled workflows, multi-step automationFully automated
AgentsContext-aware, intelligent automationAI-driven
MCP serverExternal AI system integrationAI-driven
APICustom applications, programmatic accessFully automated