Tools provided
Tools the agent exposes — your AI client calls them automatically when it needs them.
- run_full — Run the POS Sales Velocity Analyzer end-to-end: collect Square POS data, apply the Menu Engineering matrix, synthesize an LLM narrative, return the full AgentResult JSON.
- analyze_menu — Given a raw menu dataset, run the deterministic Menu Engineering matrix (no LLM) and return classified items, recommendations, and findings.
- get_recommendations — Get prioritized menu recommendations filtered by ME category (dogs, puzzles, plow_horses, stars, or all).
- classify_item — Classify a single menu item into the Menu Engineering quadrant given its sales stats.
- list_capabilities — List the agent's static capabilities: available tools, ME categories, and credential slots.
- plan_inputs — Plan inputs for a tool: returns questions, schema, and a ready-to-edit example.
- discover_intent — Proactive intent discovery: restate the goal, ask clarifying questions, propose a concrete input.
