Better results for every document.

Optimize accuracy, reliability, latency, and cost across document providers through one durable API.

Start for free
npx @file_router/cli@latest login

Works with

LlamaParse
Firecrawl
Datalab
Mistral OCR

Building blocks for the pipeline you want.

Focused engines for cheap paths, hard docs, and durable jobs.

document

Source

FileRouter

Durable job

parse()compare()

ParseResult

Normalized output

Durable jobsSafe retriesStored resultsAutomatic cleanup

Same interface for every engine.

Point at liteparse or pdf-inspector for simple pages, a heavier engine when you need it. Same typed result.

import { FileRouter } from "@file_router/sdk"


const router = new FileRouter()


// Fast hosted engine for simple pages
const result = await router.parse(file, {
  provider: "liteparse",
  outputs: ["markdown", "tables"],
})

Different documents. Different winners.

No single engine wins every doc. Here's the tradeoff.

ParseBench

Quality vs cost across providers

Up and to the right is better value.

Overall vs ¢ / page · cheaper →

0.50¢1.5¢6080OverallAverage cost per pageLlamaParse Cost EffectiveLlamaParse AgenticDatalab AccurateDatalab BalancedDatalab FastMistral OCR 4 (Annotation)Mistral OCR 4

Scaled to supported providers (0.17¢–1.94¢/page) · 27 outliers hidden · Latest public leaderboard · accessed July 20, 2026 · 2,078 pages across insurance, finance, and government documents. Scores are 0–100; higher is better.

Simple, flexible pricing

Use your provider keys for free, or use FileRouter credits for hosted processing.

Direct (BYOK)

Call providers from your own runtime with your own keys.

$0

  • No FileRouter credits
  • Pay providers directly
  • Typed SDK and CLI
  • Parse and compare locally

Hosted

Send documents through FileRouter for hosted processing.

Pay as you go

  • FileRouter API key
  • 5,000 free credits each month
  • Durable jobs and retries
  • Uploads and cleanup included

Enterprise

Custom deployment, support, and controls for your team.

Custom

  • Custom usage and volume
  • Priority implementation support
  • Security and compliance review
  • Custom deployment options

Assemble the pipeline. We run the jobs.

Pick the engines that fit each document. FileRouter handles durable execution, retries, results, and cleanup.