Configuration¶
The module has no settings form. Everything is configured per node in the FlowDrop editor; see AI Agent Executor node.
Permissions¶
The module defines no permissions of its own. It relies on these:
| Permission | Provided by | What it grants here |
|---|---|---|
administer flowdrop |
FlowDrop | Access to the agent and search index endpoints under /api/flowdrop/, which the agent selector widget calls. Without it the selector cannot list agents. |
administer ai_agent |
AI Agents | Shows an agent's system prompt in the selector preview. Users without it see the label, description and tools only. |
The agent detail endpoint also checks view access on the ai_agent config entity, so access
hooks on agents apply there too.
Agents act with real write access
Agents can create, edit and delete configuration and content through their tools. A workflow
runs the agent with whatever access the agent's tools allow. Review which agents a workflow
calls and set create_directly to false where you want a blueprint instead of changes.
Node type¶
The node type is a FlowDrop config entity, flowdrop_node_type.flowdrop_agents_ai_agents_executor.
It marks each parameter as configurable (set in the node form), connectable (wired from another
node) or both:
| Parameter | Configurable | Connectable | Required |
|---|---|---|---|
agent_id |
yes | no | yes |
task_description |
no | yes | yes |
task_title |
yes | no | no |
chat_history |
no | yes | no |
create_directly |
yes | no | no |
additional_context |
yes | yes | no |
tool_usage_limits |
no | yes | no |
llm_model |
no | yes | no |
Outputs information and execution_metadata are not exposed as ports by default. You can change
these flags by editing the node type config, as with any FlowDrop node type.