Aakash Harish, Ameya, Piyussh, and Chaitanya Lambat, Singhal, and Chaurasia
Pre-Seed
United States
AI, SaaS, B2B, Enterprise
2026
Open core. The rightmodeler skill is free and MIT licensed, runs entirely on the customer's own provider key, and is the wedge: it earns developer trust and teaches us every trace format in the wild. Two paid products sit above it. rightmodeler agent watches each new model release, replays it against the team's own traces in CI, and opens a pull request with the evidence attached, sold per watched pipeline per month. Crucible, the analytics and continuous right-sizing suite for agent fleets, is sold per fleet, with an enterprise tier for SSO, self-hosting, and audit. We sell to the engineering leader at an AI product company who owns the production inference bill. Revenue recurs by construction: a model choice decays as prices fall and new models ship, so staying current is a subscription-shaped problem and every model launch reactivates the account. The company is still in stealth: the open-source skill is public, but we have not launched publicly, no price is published yet, and both paid products are on waitlist and early access.
Still in stealth. rightmodeler decides which model belongs at every layer of your AI agents: it replays your own traces through cheaper candidates, judges each output against what you already shipped, and reports the evidence. Stop paying frontier prices for work a cheaper model does as well.