Private preview

Probability. Confidence. Evidence.

Calibrated probabilities your agents can reason about. The evidence behind every call. The same record the dashboard quotes, exposed as four HTTP routes.

01 · The Audience

Built for agents and enterprise integrators.

The same calibrated forecast a person sees in the dashboard. Every response carries a probability, the evidence behind it, and a confidence interval an agent can reason about.

02 · Endpoints

Four routes. One record.

Every response carries the probability, the confidence, and the evidence behind it.

01

GET/v1/forecast

Returns the current probability for a forecast subject, with the evidence chain attached. Which inputs moved, when, and by how much.

02

GET/v1/registry

Browse resolved forecasts with filters by subject, region, horizon, and date. Every entry has a resolution, a score, and the call’s history.

03

POST/v1/scenarios

Submit a scenario spec. Receive a probability distribution across branching outcomes. Useful for what-if reasoning before commitment.

04

GET/v1/calibration

Returns the public track record for any subject, region, or horizon. Machine-readable. The same record the dashboard quotes.

03 · Access

Preview to production. Four steps.

Each step is operator-led. Token rotation per your policy.

01 · Request

Tell us what you’re building.

Submit your use case, expected request volume, and the system you want to embed Meridian into.

02 · Authenticate

Receive a bearer token.

Token rotation handled per your policy. Rate limits and quota posted with the docs.

03 · Integrate

Hit the endpoints.

Any HTTP client works. Responses are typed JSON. The same forecast a person sees in the dashboard.

04 · Production

Move from preview to production.

Once your evaluation cycle closes. Every endpoint returns the evidence behind the probability, so an agent can reason about the result.

04 · The case

Why an API.

Agents need probabilities they can act on, not paragraphs of caveats. A narrative confidence does not survive a chain of model calls.

Meridian returns probability, confidence, and evidence behind every call. Same model as the dashboard. Same public record. Different surface.

The track record

Every call sits on the same public scale.

Hold-out backtests against decades of history. Brier scores rolling thirty days. Reliability decomposition machine-readable through /v1/calibration. The record the dashboard quotes is the same record the API exposes.

05 · Next steps

Request developer access.

Tell us what you want to embed Meridian into. We’ll get you a token and the early docs.

Request developer access