Freight-document extraction inside your own workflow.

Your first document

1. 1

   Create a key

   Scale and Custom workspaces can create a Process + read key in Dashboard → API.

2. 2

   Submit the PDF

   Send multipart form data with a unique Idempotency-Key for each logical document.

3. 3

   Get the result

   Poll the document URL or wait for document.completed, then fetch JSON or Excel.

curl -X POST "https://manifestai.io/api/v1/documents" \
  -H "Authorization: Bearer $MANIFEST_AI_API_KEY" \
  -H "Idempotency-Key: invoice-acme-2026-1042" \
  -F "file=@commercial-invoice.pdf" \
  -F "profile_id=YOUR_PROFILE_ID"

import os
import requests

with open("commercial-invoice.pdf", "rb") as document:
    response = requests.post(
        "https://manifestai.io/api/v1/documents",
        headers={
            "Authorization": f"Bearer {os.environ['MANIFEST_AI_API_KEY']}",
            "Idempotency-Key": "invoice-acme-2026-1042",
        },
        files={"file": ("commercial-invoice.pdf", document, "application/pdf")},
        data={"profile_id": "YOUR_PROFILE_ID"},
        timeout=60,
    )

response.raise_for_status()
print(response.json()["data"]["id"])

import { openAsBlob } from "node:fs";

const form = new FormData();
form.set("file", await openAsBlob("commercial-invoice.pdf"), "commercial-invoice.pdf");
form.set("profile_id", "YOUR_PROFILE_ID");

const response = await fetch("https://manifestai.io/api/v1/documents", {
  method: "POST",
  headers: {
    Authorization: `Bearer ${process.env.MANIFEST_AI_API_KEY}`,
    "Idempotency-Key": "invoice-acme-2026-1042",
  },
  body: form,
});

if (!response.ok) throw new Error(await response.text());
console.log((await response.json()).data.id);

Bearer keys with least privilege

Send Authorization: Bearer mai_live_…. Keys are shown once, cannot be recovered, and should be stored in a secret manager—not source code or a browser application.

Durable, credit-aware processing

API submissions use the same encrypted storage, duplicate detection, page credits, saved profiles, processing mode, retention window, and worker checkpoints as portal uploads. Every API record appears in Recent documents with an API label.

curl "https://manifestai.io/api/v1/documents/DOCUMENT_ID" \
  -H "Authorization: Bearer $MANIFEST_AI_API_KEY"

Endpoints

For complete request and response schemas, import the OpenAPI 3.1 specification.

Signed, retry-safe status events

Subscribe to document.completed, document.failed, and document.cancelled. Manifest AI sends metadata and resource links—not extracted customer fields. Retrieve the result using your API key.

const signed = timestamp + "." + rawRequestBody;
const expected = hmacSha256(WEBHOOK_SECRET, signed);

// Reject stale timestamps, then compare signatures
// with a timing-safe equality function.

• HTTPS endpoints only; private or local network targets are blocked.
• No redirects are followed during delivery.
• Verify Manifest-Timestamp and Manifest-Signature against the raw body.
• Return any 2xx response quickly; failed deliveries retry with backoff.
• Use Manifest-Event-Id to deduplicate deliveries.
• Rotate application API keys independently from webhook signing secrets.

Errors, limits, and retries

Every JSON response includes X-Request-Id. Authenticated responses also expose X-RateLimit-Limit, X-RateLimit-Remaining, and X-RateLimit-Reset. HTTP 429 includes Retry-After.

{
  "error": {
    "code": "DOCUMENT_NOT_FOUND",
    "message": "Document not found.",
    "request_id": "a74b0a33-..."
  }
}