Introduction

Steve is an API for document processing, AI extraction, fraud detection, and submission review.

Steve is a headless document processing platform. Integrators submit files through an API, Steve runs AI extraction and fraud detection, and the results land in a review queue — ready for approval, rejection, or automated downstream sync.

How it works

Discover — list workflows your company can use and fetch the upload contract
Upload — create a session, receive pre-signed R2 URLs, upload files directly to Cloudflare R2
Submit — queue the session for asynchronous processing
Track — receive webhook callbacks or poll the job endpoint
Review — inspect results, resolve fraud matches, approve or reject

Start here

Quick Start

Submit your first file through the API in 5 minutes.

Sandbox

Try the API in a hosted demo environment — no setup required.

API Reference

Request and response schemas for every endpoint.

API surface

Surface	Audience	Authentication	Docs
External Integration API	Server-to-server integrations	`Authorization: Bearer aok_...`	API Reference
Invitation API	Browser onboarding	Invitation token	Internal
Workflow Agent API	Internal maintainers	Convex session bearer + `super_admin`	Internal

Key concepts

Workflows define what Steve processes — accepted file types, extraction schema, output contract
Sessions are short-lived upload windows with pre-signed R2 URLs (30 min TTL)
Submissions are the review unit — they carry extracted data, fraud matches, and event history
API keys are company-scoped, long-lived server secrets prefixed aok_
All responses use camelCase JSON, ISO 8601 timestamps, and RFC 9457 error envelopes

Base URL

https://<deployment>.convex.site

If your app stores a Convex client URL like https://<deployment>.convex.cloud, replace .cloud with .site. See Base URLs.