Tools provided
Tools the agent exposes — your AI client calls them automatically when it needs them.
- run_full — Orchestrated path: translate a natural-language product question into a funnel query, run it over your funnel snapshot, and return the full funnel with drop-off percentages, segment breakdown, anomaly note, findings and a narrated hypothesis. Without data it returns a clearly labelled demo fixture.
- analyze — Run pure deterministic funnel and anomaly analysis over a provided snapshot with no model call, returning scored findings.
- compute_conversion — Compute the overall conversion — last step divided by first step — with the first and last user counts. Deterministic, no model call.
- find_biggest_dropoff — Return the single worst non-entry step: the biggest drop-off in the provided funnel. Deterministic, no model call.
- detect_anomaly — Return the flagged anomaly and its scored finding, comparing actual drop-off against expected. Deterministic, no model call.
- rank_step_findings — Return the per-step drop-off findings, worst first. Deterministic, no model call.
- breakdown_segments — Return the segment drop-off breakdown, worst first — which platform, country or cohort is dragging the funnel. Deterministic, no model call.
- list_capabilities — List the agent's static capabilities: the available tools and the analyzer's drop-off and anomaly thresholds. Useful for discovery.
- plan_inputs — Plan the inputs for a tool: returns the clarifying questions, the JSON schema and a ready-to-edit example.
- discover_intent — Understand your goal and co-design the exact input through clarifying questions before running.
- open_form — Opens the guided form to collect the question, the funnel snapshot, the date range and the segment. The entry point for running the agent on your own data.
- run_form — Internal: invoked by the guided form when the user submits. Not for direct use.
