Our biggest upgrade yet · free while we build it

We rebuilt the reasoning engine from scratch. Now it's free because we need your feedback.

Version 2.0 ships with a custom 340M-parameter model trained on a cluster of NVIDIA A100 GPUs. Retrieval is 10× faster, recall@10 jumped from 68% to 94%, and every answer still carries its full six-step trace. We're giving it away during beta so real users can stress it, break it, and tell us what to fix before we go to general availability.

340M new model params

180 ms p50 latency

94% recall@10

A100 training cluster

Why is v2.0 free right now?

Real-world feedback. Your queries find the edge cases our internal tests miss.
Faster bug fixing. Every reported issue goes straight to the engineering team in Colombo.
Shape the roadmap. What you ask for now is what we ship in 2.1 and 2.2.
No credit card, no lock-in. Starter plan is free forever. Upgrade only when you're ready.

Deterministic 6-step reasoning chain
Per-step confidence, calibrated to a Brier score of 0.08
Interactive entity graph with shortest-path search
Immutable audit trail on every action

You're early. Help us get this right.

Every workspace gets a direct feedback line to the team. Found a bug? Tell us. Missing a feature? Ask. We ship weekly.

Create your account

Free during the v2.0 beta period. No credit card.

Full reasoning engine

No credit card

Direct line to the team

Founder pricing at GA

Full name Work email Password

Use 8+ characters with a mix of letters, numbers and symbols.

Confirm password

I agree to the terms of service and privacy policy.

By creating an account you join the v2.0 beta. We'll email you occasionally about fixes and new features nothing else.

Welcome back

Create your account