What it does
Funnel Cohort Analyst computes funnel step conversion and cohort retention from caller-provided event data. It is account-free: there is no warehouse connector and no credentials — you pass events in as data, and the agent does the rest. Given a stream of events (userId, step, timestamp, optional cohort), it computes overall conversion, step-over-step conversion, and drop-off for each funnel stage, plus a full cohort retention matrix at day, week, or month granularity. All metrics are produced by a pure deterministic core, so the same events always yield the same numbers. A language model is used only to narrate and prioritize the computed figures — it can never invent or alter a number — and if no LLM is available, the agent falls back to a deterministic template so it always returns a useful result. Privacy and safety are built in: user IDs are used only as aggregation keys and never echoed into findings, prompts, or output; only aggregated numbers ever reach the LLM; event fields are sanitized so prompt-injection hidden in the data cannot reach the model as an instruction; and a network egress sandbox is enabled by default since the agent needs no external network. Designed for product and data teams who want fast, repeatable, privacy-safe funnel and retention analysis without wiring up a warehouse connector or sharing raw user-level data.
Tools (3)
Tools the agent exposes — your AI client calls them automatically when it needs them.
- run_full — Run the full funnel + cohort analysis on caller-provided events end-to-end: deterministic funnel conversion + cohort retention, then an LLM narrative of the numbers (template fallback). Returns the complete AgentResult JSON.
- analyze_funnel — Compute the funnel conversion + cohort retention matrix from caller-provided events using pure deterministic math (no LLM). Returns funnel, cohorts, findings and totals.
- list_capabilities — List the agent's static capabilities: available tools and analysis parameters. Useful for discovery.
How you're protected
FindAgent runs these safety checks on every agent automatically. They're always on and can't be turned off.
- Prompt-injection scanning
Every request is checked for known prompt-injection and jailbreak attempts before the agent runs. This is always on.
- Secret-leak scanning
Every response is scanned for leaked API keys, tokens, and other secrets before it reaches you. This is always on.
