API Reference

The predictAI API is a JSON HTTP API under a single versioned prefix, /v1. This page covers the conventions that apply everywhere; each product section documents its own endpoints in depth.

Base URL

https://api.predict.ai

One hostname works for every account. Your data lives in a home region, and if a request lands elsewhere the API answers 409 with the correct base URL in error.details.redirect_api_base_url — the official SDKs follow that redirect automatically and cache it, so you never have to think about regions. If you'd rather skip the one-time redirect hop, use the exact base URL shown next to your API token in the app. See Authentication → Regions.

Headers

curl "$API_BASE/v1/signal" \
  -H "Authorization: Bearer pa_live_..." \
  -H "X-Workspace-Id: $WORKSPACE_ID" \
  -H "Content-Type: application/json"

from predictai import PredictAI

client = PredictAI(token="pa_live_…", workspace_id="ws_…")

data = client.signals.get_signal()

import { PredictAI } from "@predictai/sdk";

const client = new PredictAI({ token: "pa_live_…", workspaceId: "ws_…" });

const data = await client.signals.getSignal();

Conventions

• Envelopes — most responses wrap payloads in {"data": …}; acknowledgements use {"status": …}; inference and deployments return top-level payloads. See Errors & responses.
• Timestamps — ISO 8601, UTC (2026-07-15T00:00:00Z).
• IDs — opaque strings. Don't parse them.
• Pagination — list endpoints take page / per_page (or limit / offset) query params and return totals alongside the items.
• Async work — long operations return 202 or a queued status with an ID to poll; prefer Realtime subscriptions over polling.

Endpoints by resource

Platform endpoints

Unauthenticated basics:

Machine-readable spec

The OpenAPI 3 snapshot for the public API is served at /openapi/v1.json — use it for codegen, Postman, or AI tooling. An AI-friendly index of these docs lives at /llms.txt.