llms.txt
# Reducto
> Reducto is the complete agentic document platform for leading AI teams that need performance at enterprise scale. Document work starts here: the most accurate document processing available — parsing, extraction, splitting, and classification — plus document editing, workflows, and agent-ready tooling, in one platform.
Reducto provides a comprehensive toolkit for working with documents the way a human would, combining custom in-house models with leading frontier models to power efficient, accurate document workflows. Teams replacing brittle OCR pipelines and legacy intelligent document processing (IDP) tools choose Reducto for zero-shot accuracy on complex documents — tables, charts, figures, handwriting, scans — across 30+ file types. Over 5 billion pages processed for AI teams at companies like Harvey, Scale AI, and Vanta.
Get started free with 15,000 credits. Agents can create a new account through HTTP; existing users and companies continue in [Studio](https://studio.reducto.ai). Read the [authentication runbook](https://reducto.ai/auth.md) and [Agent skill](https://reducto.ai/SKILL.md).
## Setup for an assistant
1. Check a saved REDUCTO_API_KEY with GET https://platform.reducto.ai/jobs?limit=1 and a bearer header. Never print the key. Use Studio to recover an invalid key.
2. With no key, get the human's email. POST https://accounts.reducto.ai/api/agent/accounts with Content-Type: application/json and {"email":"ada@example.com"}. Save the returned signup_token privately; the code expires in 10 minutes. No account is created yet.
3. Ask for the six-digit emailed code, first/last name, explicit agreement to https://reducto.ai/terms and acknowledgement of https://reducto.ai/privacy, and the first document together. Explain that sharing the code authorizes account creation and an API key for the assistant.
4. POST https://accounts.reducto.ai/api/agent/accounts/verify with {"signup_token":"<saved token>","code":"012345","first_name":"Ada","last_name":"Lovelace","terms_accepted":true}. On 202 {"status":"provisioning","retry_after":5}, wait and repeat the identical request. Never restart signup during provisioning.
5. On 200, save api_key as REDUCTO_API_KEY in the approved secret store; the response also includes api_key_id, org_id, credits_granted:15000, expires_at, and studio_url. Never paste the key or full response into chat. Identical verification requests recover the same result for 10 minutes, then use Studio. After successful signup, tell the human: no password was set; the signup code only verified their email. They can sign into Studio using a magic link via “Email me a sign-in link”. Setting a password is optional via “Set a password” on the login screen.
6. Upload the human's document with POST https://platform.reducto.ai/upload (multipart file), then POST /parse with {"input":"<returned file_id>"} and the bearer key. Share the result and studio_link.
Show validation errors in the terminal. For 400 invalid_input, display error.message and error.field and ask for corrected input; error.restart_required means start a new code request after corrections and the resend cooldown. Otherwise preserve already accepted verification data. For 403 signup_restricted, display the policy message and ask for a different email. Only 409 studio_required requires Studio. Honor Retry-After on 429/503; 400 invalid_code means ask the human, 410 verification_expired means request a new code after the cooldown, and 409 studio_required means stop automated signup and show error.message with studio_url. The human continues in Studio to sign in or select Join or Request access for a listed organization. See [auth.md](https://reducto.ai/auth.md) for exact response shapes and every error. Signup OpenAPI: https://accounts.reducto.ai/api/agent/openapi.json. Full processing documentation: https://docs.reducto.ai/llms.txt.
Agent skills discovery: https://reducto.ai/.well-known/agent-skills/index.json
## Products
- [Parse](https://reducto.ai/parse): The ingestion layer. Turn PDFs, scans, spreadsheets, and slides into structured, citation-ready JSON for LLM, RAG, and agent workloads at production scale. API docs: https://docs.reducto.ai/parse/overview.md. For choosing a parser, see [Best Document Parsers for RAG](https://reducto.ai/guides/best-document-parsers-rag).
- [Split](https://reducto.ai/split): The segmentation layer. Describe sections in plain language; Split returns the page ranges for each. Partition keys group repeating sub-documents automatically. API docs: https://docs.reducto.ai/split.md. For where segmentation fits in a larger pipeline, see [Intelligent Document Processing](https://reducto.ai/guides/intelligent-document-processing).
- [Extract](https://reducto.ai/extract): The extraction layer. Return defined fields as typed JSON with a citation and bounding box on every value. Schemas, prompts, and Deep Extract for production accuracy on the long tail. API docs: https://docs.reducto.ai/extract/overview.md. For comparing extraction options before you build, see [Best Enterprise Data Extraction Software](https://reducto.ai/guides/best-enterprise-data-extraction-software).
- [Classify](https://reducto.ai/classify): The routing layer. Define your taxonomy in plain language and Classify returns the best match plus per-criterion confidence, fast enough to gate every pipeline. API docs: https://docs.reducto.ai/classify/overview.md. For how classification platforms differ on routing and human review, see [Best Document Classification Software](https://reducto.ai/guides/best-document-classification-software-2026).
- [Edit](https://reducto.ai/edit): Closes the agentic document loop, turning extracted data back into a finished file. Vision-based field detection, reusable form schemas, and optional DOCX highlighting at production scale. API docs: https://docs.reducto.ai/editing/edit-overview.md. For the architecture around write-back steps, see [Document Workflow Automation](https://reducto.ai/guides/document-workflow-automation).
- [Studio](https://reducto.ai/studio): The no-code workspace for the platform. Build, test, and deploy document pipelines without code — same engine as the API, included with every plan. Docs: https://docs.reducto.ai/studio-quickstart.md. For scoring platforms against enterprise requirements, see [Enterprise IDP Evaluation Guide](https://reducto.ai/guides/idp-enterprise-evaluation).
## Key Differentiators
- **Accuracy where others aren't production ready**: zero-shot accuracy on complex documents; Deep Extract achieves 99% recall and precision on micro1's LongExtractionBench.
- **12+ models orchestrated**: automatic routing balances accuracy, latency, and throughput for each document's complications, continuously updated so your team doesn't chase the frontier. Reducto uses the best parts of frontier models rather than competing with them.
- **Long-tail robustness**: complex tables, charts, figures, handwriting, scans, checkboxes, equations, rotated pages, and multilingual documents — without templates, pre-labeling, or retraining when documents change.
- **Trustable outputs**: a citation and bounding box on every extracted value, with a side-by-side citation viewer in Studio.
- **One platform, every document task**: parse, split, extract, classify, and edit across 30+ file types — from raw file ingestion to agent-ready outputs and workflow orchestration, replacing 4–5 stitched-together vendors.
## Enterprise Readiness
- **Deployable anywhere**: hosted cloud, hybrid VPC, on-premises, and fully air-gapped to meet any data residency or security requirement
- **Compliant**: SOC 2 and HIPAA compliant (BAAs on Growth tier and above), zero data retention agreements on Growth tier and above
- **Production scale**: autoscaling built in for bursty workloads, white-glove forward-deployed engineering support, custom SLAs — over 5 billion pages processed
- Details: https://docs.reducto.ai/security/policies
## Industries & Use Cases
- [Finance](https://reducto.ai/industries/finance): Research reports, SEC filings, financial statements, and complex spreadsheets with decimal-level fidelity
- [Healthcare](https://reducto.ai/industries/healthcare): HIPAA-compliant processing of medical records, clinical notes, and claims
- [Legal](https://reducto.ai/industries/legal): Contract review, due diligence, NDAs, and filings with citation-ready accuracy
- [Insurance](https://reducto.ai/industries/insurance): Underwriting, claims, and ACORD forms on mixed-format evidence with audit-ready accuracy
- [Government & Defense](https://reducto.ai/industries/government): Secure processing of forms, permits, intelligence reports, and regulatory filings
- [Construction](https://reducto.ai/industries/construction): Blueprints, contracts, bids, invoices, and submittals
- [Logistics & Supply Chain](https://reducto.ai/industries/logistics-supply-chain): Bills of lading, customs forms, invoices, and shipping manifests
Cross-industry use cases: invoice and AP automation, KYC and onboarding, RAG preprocessing and embedding-optimized chunking, replacing manual data entry, and document generation workflows. Full case studies at https://reducto.ai/customers.
## Developer Tools
- [Agent skill](https://reducto.ai/SKILL.md): Self-contained SKILL.md for AI agents — setup, API usage, CLI, and MCP server instructions
- [Agent card](https://reducto.ai/.well-known/agent-card.json): Machine-readable agent card for Reducto's MCP and REST API interface
- [Agent discovery](https://reducto.ai/.well-known/agents.json): Discovery metadata linking Reducto's agent card, skill, and MCP endpoint
- [MCP discovery](https://reducto.ai/.well-known/mcp.json): Machine-readable MCP endpoint, authentication, and server-card metadata
- [Web MCP discovery](https://reducto.ai/.well-known/webmcp.json): Machine-readable Web MCP endpoint and server-card metadata
- [MCP server cards](https://reducto.ai/.well-known/mcp/server-cards.json): Index of Reducto's hosted MCP server cards
- [MCP server](https://docs.reducto.ai/mcp-server): Connect AI agents directly to Reducto via the Model Context Protocol. Hosted endpoint at https://mcp.reducto.ai/mcp (Bearer API key) or local via `uvx mcp-server-reducto`. Works with Claude Code, Claude Desktop, Codex, Cursor, VS Code, Windsurf, and any MCP client.
- [CLI](https://docs.reducto.ai/cli): `pip install reducto-cli` — parse, extract, split, classify, and edit from the terminal. Built for batch processing, scripting, and CI/CD; writes agent-readable Markdown next to each input file.
- [Agent integration guide](https://docs.reducto.ai/agent-guide.md): Complete, structured API reference written for AI coding agents
- [API documentation](https://docs.reducto.ai): Quickstarts, API reference, and SDKs for Python and Node. For the developer-side comparison of PDF extraction APIs and SDKs, see [Best PDF Data Extraction APIs and SDKs](https://reducto.ai/guides/best-pdf-data-extraction-apis-sdks)
- [OpenAPI specification](https://reducto.ai/openapi.json): OpenAPI 3.1 schema for the Reducto API
- [Documentation index for LLMs](https://docs.reducto.ai/llms.txt): Curated index of all docs pages, each available as raw Markdown
- [Checking API Health & Usage](https://docs.reducto.ai/reference/checking-api-health.md): How to check hosted API availability (via https://status.reducto.ai), usage, and throttling signals
- [Agent capabilities catalog](https://reducto.ai/.well-known/ai-catalog.json): Machine-readable list of Reducto's MCP servers
- [API catalog](https://reducto.ai/.well-known/api-catalog): RFC 9727 machine-readable index of Reducto's APIs, including the OpenAPI specification
## Get Started
- [Reducto Studio](https://studio.reducto.ai/): Sign up and get 15,000 free credits to try the platform on your own documents
- [Pricing](https://reducto.ai/pricing): Per-product plans — Standard (self-serve pay-as-you-go with 15,000 free credits), Growth (custom volume pricing with security and compliance features), and Enterprise (custom pricing with VPC, on-premises, and air-gapped deployment). r-1 Parse starts at $10 per 1,000 pages, and Extract starts at $20 per 1,000 pages with parsing included; Markdown: https://reducto.ai/pricing.md
- [Contact](https://reducto.ai/contact): Talk to sales or request a demo; Markdown: https://reducto.ai/contact.md
- [About](https://reducto.ai/about): Company, products, customers, compliance, and contact details; Markdown: https://reducto.ai/about.md
- [Developers](https://reducto.ai/developers): API, SDK, CLI, and MCP implementation guidance; Markdown: https://reducto.ai/developers.md
- [Authentication](https://reducto.ai/auth.md): How agents obtain and use Reducto API credentials
## Company
- [Customers](https://reducto.ai/customers): Case studies from AI teams running Reducto in production, including Harvey, Scale AI, and Vanta
- [Industries](https://reducto.ai/industries): Industry solutions hub
- [Blog](https://reducto.ai/blog): Technical articles, case studies, and product announcements
- [Engineering blog](https://reducto.ai/engineering/llms.txt): Deep dives from the team building Reducto, indexed for LLMs
- [Careers](https://reducto.ai/careers): Open positions
- [Reducto for Startups](https://reducto.ai/startups): Credits, startup pricing, and priority support for teams scaling document workflows from prototype to production
- [Reducto for YC](https://reducto.ai/yc): The exclusive YC deal — startup pricing and credits for YC companies, from the team that went through YC W24
## Optional
- [Cookbooks](https://reducto.ai/cookbooks): Templates and recipes for building with Reducto
- [Guides](https://reducto.ai/guides): Buyer guides, comparisons, benchmarks, and technical explainers for document AI
- [Compare](https://reducto.ai/compare): Tool-by-tool comparisons of Reducto vs other document parsing, OCR, and extraction options. For the published evidence behind those comparisons, see [Document AI Benchmark: OCR, Parsing and Extraction](https://reducto.ai/guides/document-ai-benchmark-ocr-parsing-extraction)
- [Privacy policy](https://reducto.ai/privacy)
- [Terms of service](https://reducto.ai/terms)
