Execute a single automation node directly (without a workflow graph).
This invokes any registered node handler: the same code that runs inside an automation workflow, but as a standalone one-shot call. Use automations_list_node_types to discover available node types and their expected config keys.
Args:
node_type: The node type identifier, e.g. 'function_create_task',
'action_slack_send_message', 'function_http_request',
'action_openai_chat_completion', 'query_task', etc.
config: Node-specific configuration dict. Keys vary by node type.
For API keys and passwords, **never put raw secret values here**.
Call automations_list_secrets first, then reference a secret
by id::{"anthropic_api_key": {"$secret": "<uuid>", "_label": "My key"}}
The server resolves markers at execution time using the
caller's access grants; values are redacted from responses.Common examples:
**function_create_task**::
{"title": "...", "description": "...", "widget_id": "...",
"status": "...", "assignee_id": "..."}**function_http_request**::
{"url": "https://...", "method": "GET",
"headers": [{"name": "Authorization", "value": {"$secret": "<uuid>", "_label": "token"}}]}**seo.fetch_keyword_universe**::
{"dataforseo_login": {"$secret": "<uuid>", "_label": "DF login"},
"anthropic_api_key": {"$secret": "<uuid>", "_label": "Claude"}}**action_openai_chat_completion**::
{"model": "gpt-4o", "messages": [{"role": "user", "content": "..."}]}
**action_slack_send_message**::
{"channel": "#general", "text": "Hello from automations"}
**query_task**::
{"task_id": "abc123"}
Values support {{nodes["nodeId"].field}} template syntax
for referencing prior node outputs (in workflow context).
space_id: Productivity workspace/space ID for context. Required
for productivity nodes (task/batch operations). Optional
for global nodes (HTTP, AI, etc.).Returns the node handler's output as JSON. Error responses include an error key with a description.
Long-running worker nodes (the seo.*, research.*, and editorial video.* families can take several minutes of external API + LLM passes) support async_mode: true in config: the call returns a job_id immediately and the pack's <prefix>.get_job_result node (seo.get_job_result, research.get_job_result, video.get_job_result) fetches the outcome. Prefer that over holding this call open when a run may exceed a few minutes.