Parameter Configuration

Note: Custom Actions require a Builder plan or higher. Upgrade to Builder

Headers#

HTTP headers sent with the request.

{
  "Authorization": "Bearer {{var.API_KEY}}",
  "Content-Type": "application/json",
  "X-User-ID": "{{system.user_id}}"
}

Query parameters#

URL parameters appended to your endpoint.

GET https://api.example.com/rooms?site={{site}}&limit=10&page={{page}}

Body parameters#

JSON data sent on POST, PUT, and PATCH requests.

{
  "site": "{{siteCode}}",
  "room": "{{roomName}}",
  "metadata": {
    "source": "avcodex",
    "timestamp": "{{system.timestamp}}"
  }
}

Path parameters#

Dynamic URL segments replaced at runtime.

https://api.example.com/sites/{{siteId}}/rooms/{{roomId}}

https://api.example.com/sites/123/rooms/456

Fixed value#

Static values that never change.

Use for: API versions, static identifiers, constant headers.

Let AI generate#

Dynamic values the agent fills in from the conversation.

Use for: search queries, dynamic filters, contextual values like a site code or device hostname.

System variables#

Platform-provided runtime values.

Use for: user context, audit trails, conversation analysis.

Application variables#

Reusable values defined at the agent level.

Use for: API keys, base URLs, workspace IDs.

Output from another Action#

Use a previous action's response as input to the next one. This is how multi-step workflows work, with data flowing between API calls.

Configuration fields

How it works

1. The prerequisite action runs first and returns data.
2. The JSONPath selector extracts the needed value (or the agent uses the full response if blank).
3. The extracted value replaces the parameter in the current action.
4. If the prerequisite has not run yet, the agent runs it first automatically.

JSONPath examples

$.id                    // Top-level field: {"id": "123"} → "123"
$.data.user.email       // Nested field: {"data": {"user": {"email": "a@b.com"}}} → "a@b.com"
$.items[0].name         // First array item: {"items": [{"name": "First"}]} → "First"
$.items[*].id           // All IDs: {"items": [{"id": 1}, {"id": 2}]} → [1, 2]
$.data.users[?(@.active)] // Filtered: active users only

Real-world example

Scenario: open a support ticket against a tech's home org.

Action 1: Get tech profile

• Returns: {"user": {"id": "u123", "org_id": "org456", "name": "John"}}.

Action 2: Open ticket

• Parameter: organization_id.
• Source: Output from another Action.
• Action: Get tech profile.
• JSONPath: $.user.org_id.
• Result: parameter receives "org456".

Important: JSONPath is optional

You have two options for extracting data:

1. With JSONPath (precise): specify an exact path like $.data.id to pull a specific value.
2. Without JSONPath (agent-driven): leave it blank and use the Context field to describe what you need. The agent finds the right value in the response.

Example without JSONPath

• Action returns: {"user": {"id": "123", "email": "tech@integrator.com"}, "org": {"id": "456"}}.
• Context: "Use the organization ID from the response."
• Result: the agent extracts "456".

Tips

• JSONPath is optional. Context alone often works.
• Test JSONPath at jsonpath.com with sample data when you need precision.
• Actions with dependencies show a link icon in your action list.
• The agent picks the right value based on the context you describe.

Use for: multi-step workflows, data enrichment, sequential API operations.

String#

Text values, URLs, identifiers.

{
  "site": "1200-market",
  "url": "https://example.com"
}

Number#

Integers and decimals.

{
  "rack_count": 12,
  "labor_rate": 185.00
}

Boolean#

True or false.

{
  "active": true,
  "commissioned": false
}

Object#

Nested JSON.

{
  "metadata": {
    "tags": ["boardroom", "priority"],
    "properties": {
      "manufacturer": "Crestron"
    }
  }
}

Array#

Lists of values.

{
  "tags": ["q-sys", "biamp", "extron"],
  "ids": [1, 2, 3]
}

Writing effective instructions#

Good example

Generate a one-line ticket subject (max 100 chars) describing the device, the room, and the symptom. Include the site code if mentioned.

Bad example

Make a subject

Providing sample values#

Sample values teach the agent about format expectations:

Required vs optional#

• Required: the agent must provide a value or the action fails.
• Optional: the agent includes the value only when relevant.

Nested parameters#

Configure complex JSON structures:

response_format.type = "json_schema"
response_format.json_schema.strict = true
response_format.json_schema.name = "analyze"

{
  "response_format": {
    "type": "json_schema",
    "json_schema": {
      "strict": true,
      "name": "analyze"
    }
  }
}

Mixed sources#

Combine sources in one action:

{
  "api_key": "{{var.API_KEY}}",       // Variable
  "user_id": "{{system.user_id}}",    // System
  "query": "{{searchTerm}}",          // AI generated
  "limit": 10,                        // Fixed
  "org_id": "{{previousAction.org_id}}" // Dependency
}

Conditional parameters#

The agent decides whether to include optional parameters:

{
  "filter": {
    "source": "Let AI generate",
    "required": false,
    "instructions": "Include only if the tech specified a manufacturer or room type"
  }
}

1. Use the built-in tester#

Preview how parameters resolve:

URL: https://api.example.com/{{endpoint}}
Query: search={{query}}

endpoint: "rooms"
query: "boardroom-4"

GET https://api.example.com/rooms?search=boardroom-4

2. Verify variable resolution#

Check that all variables resolve correctly:

• Application variables show actual values.
• System variables show placeholders.
• AI parameters show example values.

3. Test edge cases#

• Empty optional parameters.
• Special characters in values.
• Large nested objects.
• Array parameters.

• Manage variables
• Test your actions
• View examples

*AVCodex · Your AV expertise. Amplified by AI.*