The Manifold API.
Ask the geometry directly. Position, next step and field-level questions over HTTP, metered by how much of the manifold each question uses, not by seat or by minute.
Three depths of question
From a point lookup to the whole field.
Position
“Where is this learner? How sure are we?”
- You get back
- Coordinates and uncertainty
- Built for
- Every turn of a tutor or app
Local
“What’s the best next step? Is this learner drifting?”
- You get back
- Gradient, divergence, a recommended move
- Built for
- Adaptive loops
Field
“If this outcome is mastered, what else moves? What’s the path from here to there?”
- You get back
- Jacobian, geodesic, entropy across a region
- Built for
- Curriculum design, publishers, analytics
What a call looks like
Evidence in, geometry out.
Illustrative. The final interface is published with early access.
POST /v1/locate
{ "learner": "lrn_…", "evidence": [ { "outcome": "…", "result": 0.8 } ] }
POST /v1/next
{ "learner": "lrn_…", "depth": "local" }
POST /v1/geodesic
{ "learner": "lrn_…", "target": "outcome:…" }Who it's for
Learning platforms and tutoring products that want a real model of the learner underneath their language model. Publishers and assessment bodies. Later, teams building characters and companions on the emotion manifold.
Data
The API runs where your data does: on your own deployment of the manifold. Evidence goes in and geometry comes back inside your infrastructure. Nothing comes to us.
Pricing
Metered by query depth. Pricing is published with early access.