Actions37
- Agent Actions
- Chat Actions
- Datasource Actions
- Group Actions
- Language Model Actions
- Tool Actions
- User Actions
- Workspace Actions
Agent → Create Agent
AI-generatedSummary
Creates a Belake.ai agent in the specified workspace using the selected language model, system prompt, visibility settings, and optional tools or datasource. The request creates the agent as active (not a draft) and returns the API response.
Inputs
- Language Model ID (required) — Identifier of the language model the agent will use.
- Agent Name (required) — Name for the new agent.
- Agent Description (required) — Description of the agent.
- Prompt (required) — System prompt that guides the agent's behavior.
- Display Mode (required) — Choose Only Me or Groups. Only Me sends an empty group ID list; Groups requires Group IDs.
- Group IDs (required) — When Display Mode is Groups, provide a JSON array of group IDs, such as ["id1","id2"]. Invalid JSON or a non-array value causes an error.
- Tools — Optional JSON array of tool IDs to make available to the agent. Leave blank for no tools; invalid JSON or a non-array value causes an error.
- Enable Web Search (required) — Whether the agent can use web search; defaults to false.
- Enable Thinking (required) — Whether reasoning/thinking is enabled; defaults to false. When disabled, the request sends a null thinking level.
- Thinking Level (required) — Shown when Enable Thinking is on. Choose Minimal, Low, Medium, or High; defaults to Low.
- Datasource ID — Optional datasource identifier to attach to the agent. Sent only when nonempty.
- Workspace ID — Workspace identifier. The operation includes it in the request body and query string.
Output shape
A single API response item per input item, containing the created agent data.
The node sends a POST request to /v1/agents. The response is returned as JSON; if the API response is an array, each array entry becomes a separate output item. The node also sends empty file and folder ID lists and sets isDraft to false.
Examples
Example 1: Create a private agent
Provide the required name, description, prompt, language model ID, workspace ID, and other required settings. Keep Display Mode as Only Me and leave Tools and Datasource ID blank.
Example 2: Create an agent for groups
Choose Groups and enter Group IDs as a JSON array. Optionally provide tool IDs as a JSON array and a datasource ID.