# Kavara > Unidentified Data Phenomena. Already in your data. Made visible by Kirk. > Kavara is the company and platform; Kirk is its discovery engine, available today. > Start with data, discover what Kirk reveals, then investigate what it means. > Discovery and explanation are separate steps. ## Start here - [Kirk developer guide on GitHub](https://github.com/Kavara-AI/kirk-developer-guide/blob/main/START_HERE.md): primary starting point for data fit, preparation and reproducible evaluation - [Agent instructions](https://kavara.ai/agents.md): task sequence, inputs, completion criteria and runtime boundaries - [Developer entry point](https://kavara.ai/docs): preparation, connection, evidence and downstream handoff - [Structured investigations](https://kavara.ai/data/phenomena-cases.json): four synthetic cases with samples and unchanged comparisons; no model results - [Canonical catalogue](https://github.com/Kavara-AI/kirk-developer-guide/blob/main/catalogues/phenomena.json): domains, preparation profiles and exploration hypotheses - [Discovery thesis](https://github.com/Kavara-AI/kirk-developer-guide/blob/main/docs/phenomena-discovery.md): measurements, reproducibility and the distinction between discovery and application value - [MCP connection guide](https://github.com/Kavara-AI/kirk-developer-guide/blob/main/docs/connect-your-claude.md): maintained runtime setup; discover live schemas before use - [Preparation portal](https://kirk-discovery-portal.vercel.app/): local preparation and illustrative Jev handoff; no live inference - [Data Fit service](https://kavara.ai/data-fit): request a supported dataset run or ask about representation and compatibility; submission is an enquiry, not inference - [Contact](https://kavara.ai/contact): data fit and deployment enquiries Catalogue examples are Hypotheses, not validated cross-domain capabilities. Match the actual model/input contract before scoring. Consult runtime instructions for current availability, pricing and usage; do not infer them from marketing prose.