Automata Consumer ToolInstall

Summary

Analyze a user prompt (optionally extracting intents, categories, and tags), search for relevant Automata agents, validate agent relevance using an LLM if enabled, and optionally execute those agents with feedback, returning structured results including analysis, search results, relevant agents, execution outputs, and interpretation.

Inputs

• Prompt (required) — The natural language prompt provided by the calling AI agent.
• Use Analyze Prompt — Whether to analyze the prompt automatically to extract intents, categories, and tags.
• Intents (CSV) — Comma-separated intents used if prompt analysis is disabled or fallback.
• Categories (CSV) — Comma-separated categories used if prompt analysis is disabled or fallback.
• Tags (CSV) — Comma-separated tags/keywords used if prompt analysis is disabled or fallback.
• Location — Location string (city/state/country) used if prompt analysis is disabled or fallback.
• Description — Optional natural language description sent to search if prompt analysis is disabled.
• Search Limit — Maximum number of agents to return from the search (range 1-50).
• Validate Relevance with LLM — If enabled, uses an LLM to filter agents and select tasks based on relevance.
• Execute Agents — If enabled, executes the relevant agents with feedback via the Consumer SDK.
• Task — Fallback task to use when executing agents if relevance validation is off.
• LLM Provider (required) — The LLM provider to use for prompt analysis, relevance validation, and execution.
• LLM API Key (required) — API key for the selected LLM provider.
• LLM Model (required) — Model name for the LLM (e.g., gpt-4o-mini, claude-3-5-sonnet).
• Temperature — LLM temperature setting controlling response randomness (0-1).
• User Language — Preferred language for responses, adaptable to any language.
• Registry Environment — Selects Automata registry environment: dev, prod, or custom URL.
• Registry URL — Custom Automata registry URL used if registry environment is set to 'custom'.
• Options - Return Interpretation Only — If true, outputs only the interpreted meaning of results, omitting raw execution outputs.

Output shape

an object containing analysis results of the prompt, search results listing agents found, a filtered list of relevant agents (if relevance validation is enabled), execution results from running agents (if execution enabled), and an interpretation summary

The output includes multiple optional fields depending on parameter settings: if 'Return Interpretation Only' is enabled, raw execution results are omitted and only the interpretation summary is returned. Execution and relevance validation depend on user settings. The operation requires a valid LLM API key and model. Returned JSON is stringified unless consumed programmatically.