Tool Executions

​

Add a tool execution

1

Add the tool action

In the parent agent flow, select Execute tool and choose a tool.

2

Configure inputs

Use schema fields or a custom JSON payload to build inputs.

3

Test execution

Run a test execution to confirm inputs and outputs. Run this in chat to verify inputs and outputs.

​

Execute tool configuration

• Payload type: schema-driven fields or a custom JSON payload.
• Access level: full access or restricted access based on tool defaults. Use will always default with the default tool configurations set on the tool.
• Async execution: run the tool in the background, it will use a fire and forget approach.
• Context memory: store outputs into memory for agents to access, later or for context management.
• Schema validation: surface errors when filling out the payload.

​

Variable templating for tool inputs

​

What context variables are available

• Tool outputs by tool name and field path.
• Sub-agent outputs surfaced through structured outputs or raw message content.
• Runtime inputs captured earlier in the workflow, these can be files, trigger inputs, questions, chat history, or current group chat messages.
• Context variables carried across all steps these include all tool outputs for all tool execution instances and sub-agent structured outputs.

​

How variables appear in the UI

• Tools: each configured tool output is listed under the tool name.
• Runtime inputs: values captured from triggers or previous steps.
• Structured outputs: fields saved from sub-agent responses.

​

Common variable paths

• Tool output: {{ tool_key.data.field }} where tool_key is the {tool_name}_{tool_id}
• Tool output with instance: {{ tool_key__instance_id.data.field }} where tool_key is the {tool_name}_{tool_id} and instance_id is the instance id of the tool execution.
• Nested objects: {{ tool_key.data.object.child }}
• Arrays: {{ tool_key.data.items.0.sku }} gets the first item in the array.
• Structured output: {{ sub_agent_name_id.key }} where sub_agent_name_id is the {sub_agent_name}_{sub_agent_id}
• Runtime input: {{ input.question }} gets the question from the runtime input.
• Inside Loops: {{ _loop.property }} gets the property from each item when the tool is executed in a loop.

​

Variable syntax

{
  "order_id": "{{ get_order_99__1.data.id }}",
  "customer": "{{ get_order_99__1.data.customer_name }}"
}

{
  "order_id": "{{ get_order__1.data.id }}",
  "backup_id": "{{ get_order__2.data.id }}"
}

​

Nested properties

{
  "email": "{{ get_customer_99.data.customer.email }}",
  "city": "{{ get_customer_99.data.customer.address.city }}"
}

​

Filters

• default for fallbacks
• tojson to serialize objects or lists
• upper, lower, trim
• length for counts
• join for list formatting

{
  "note": "{{ order.data.note | default('No notes') }}",
  "tags": {{ order.data.tags | default([]) | tojson }},
  "email": "{{ user.data.email | lower | trim }}"
}

​

Conditionals

{
  "priority": "{% if order.data.total > 100 %}high{% else %}normal{% endif %}"
}

​

Loops

{
  "items": [
    {% for item in products.data.items %}
    { "sku": "{{ item.sku }}", "qty": {{ item.quantity }} }{% if not loop.last %},{% endif %}
    {% endfor %}
  ]
}

​

Loops and skip conditions

• Loop on: select an array path in the tool input.
• Max iterations: prevent runaway loops.
• Fail on error: stop immediately or continue on errors.
• Skip conditions: use all/any groups with nested condition sets. These can be used to skip items in a loop or to skip the tool execution if a condition is met.