Reducto
Also known as: Reducto AI, Reducto Studio
Reducto turns complex documents into structured records and processes them through reusable pipelines. Operations teams can assemble and preview a pipeline in Studio, while developers can call the same processing through the API. Parse preserves document layout, Extract returns fields with source references, and Classify and Split route different document types through processing. Edit can fill forms or modify documents. Reducto offers hosted processing and enterprise deployments in a customer's own environment.
Reducto supplies document processing for operations teams and developers building applications that read, classify or edit files. Reducto Studio provides a visual way to configure a series of steps, test those steps on sample documents and deploy a version with a stable pipeline ID. The same document processing is available through an API, a CLI and an MCP server, hosted or run locally. The platform handles more than 30 file types, including scanned PDFs, spreadsheets, presentations and files with complex tables or handwriting. Reducto has processed more than 5 billion pages.
A typical pipeline starts by identifying the type of document or splitting a file into sections. Parse preserves layout and returns blocks, tables and figures with their page positions, on either the r-1 model or the cheaper r-1 mini. Extract pulls specified values into a structured schema and can link a result back to the text it came from.
Deep Extract, Deep Split and Deep Classify run an agentic loop that checks output against the source and repeats the work where needed. Edit accepts instructions for filling a PDF form or modifying a Word document. In Studio, a domain expert can adjust extraction fields and inspect the selected region of the page before shipping the revised configuration. Engineers call the same pipeline from their application and can poll for completion or receive a webhook.
Standard usage includes $150 in free credits and up to five Studio seats, then charges by the operation and number of pages. Parse costs $1 per 1,000 pages on r-1 mini or $10 on r-1, Extract and Split $20, Deep Extract and Deep Split $40, Classify $7.50 and Edit $60, with batch jobs 20% cheaper on Parse and Extract.
Growth adds volume discounts, unlimited seats, EU and Australian residency endpoints and zero data retention, and Enterprise adds VPC and on premises deployment, role based access control and SSO with SAML. Both are quoted, and teams moving from another document processor can apply for up to $5,000 in migration credits. Reducto holds SOC 2 Type II attestation.
Its Standard terms permit Reducto to use uploaded content and outputs to train and evaluate its models, while Growth terms and signed agreements cover different data handling.
Reducto fits teams that read a variety of document layouts and need a traceable field extraction or an edited file that can continue into an existing process. Harvey relies on it for document understanding across the more than 1,300 organizations it serves, and Vanta and Drata use it to read compliance questionnaires and validate evidence. Elysian reviews insurance claims 16 times faster, Anterior turns medical records into structured data for prior authorization at more than 99% accuracy, and Finch Legal's paralegals handle five times more personal injury cases.
Stack AI customers have processed more than 5 million documents through it, and Gumloop and Benchmark build on it too. Teams can use the result citations in a review interface or route the structured records into another system they operate. A developer can store the extracted text in its own retrieval index for later use by an agent. The same workflow can stay on the hosted API or move to a private deployment where the organization controls document storage and processing.
Vendor details
Canonical URL
https://reducto.ai
Category
Enterprise operations agent
Subcategory
Document classification, extraction and editing pipelines for operations and AI engineering teams
Funding status
Reducto, Inc. announced $108 million in total funding with its Series B.
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
APIs and SDKs for Python and Node.js with a Go client for core operations, plus CLI, MCP, and enterprise storage connections for AWS, Azure, Google Cloud and Box.
In practice
An operations analyst configures a pipeline to split a packet into sections and extract the fields needed for intake. The analyst checks the source highlights in Studio before the team deploys the pipeline.
A product team building a compliance assistant passes questionnaire documents to Reducto. The assistant uses extracted fields and citations from the underlying files to prepare work for a reviewer.
A document engineering team processes scanned statements in a private deployment, then sends the structured output to its own reporting system after a completion notification.
Sources & related URLs
Related / legacy domains
Research sources
Agentic Index coverage score
8.0 / 14 capabilities · 57%
| Integrations & Tool Calling | Partial |
|---|---|
|
Reducto can process documents from storage held in the customer's cloud account. Hybrid deployments support Amazon storage, Azure Blob Storage, Google Cloud Storage and Box. Results return to the calling application, which can use them in another workflow. The named connections are concentrated in document storage and delivery rather than write actions across several business systems. Sourcedocs.reducto.ai/onprem/enterprise_deployment_optionsread 2026-11-20 |
|
| Workflow Orchestration | Full |
|
A Studio pipeline can combine parsing, splitting and extraction into one callable configuration. A split produces individual document sections, and the extraction step can return a separate structured result for each. The team deploys the configuration under a pipeline ID so a caller can run the complete sequence without invoking and coordinating each step itself. Sourcedocs.reducto.ai/workflows/pipeline-basicsread 2026-11-20 |
|
| Knowledge Grounding & RAG | Partial |
|
Extraction works from the documents the caller submits. The pipeline can turn page structure into blocks, preserve table relationships, and return selected fields with references to the underlying text. A consumer can pass the parsed material and source locations into an agent or retrieval system. An application that needs a persistent search index operates that index separately. Sourcedocs.reducto.ai/studio-quickstartread 2026-11-20 |
|
| Human Oversight & Guardrails | Partial |
|
A reviewer can inspect an extracted value next to its source in Studio and look at the highlighted location on the original page. A team can use confidence values to decide which results need attention. In production, the assignment, approval and correction workflow is owned by the customer's review process, with Reducto providing the evidence that process needs. Sourcereducto.ai/guides/human-review-document-automationread 2026-11-20 |
|
| Security, Identity & Governance | Full |
|
Reducto holds SOC 2 Type II attestation and can support HIPAA workflows on Growth and Enterprise with a business associate agreement. Enterprise organizations can configure single sign on with SAML or OIDC, and Studio manages account and team access. Reducto encrypts content in transit and at rest, and offers contractual zero data retention options on higher tiers. Sourcedocs.reducto.ai/security/policiesread 2026-11-20 |
|
| Observability & Auditability | Partial |
|
Studio shows when an API call reached a deployed pipeline, while the Activity view traces configuration changes and past versions. Extraction output can include the page location linked to each value so an operator can inspect the result. The run view covers execution and source evidence, without an end to end trace of the model's intermediate decisions. Sourcedocs.reducto.ai/workflows/pipeline-basicsread 2026-11-20 |
|
| Memory & State Persistence | Not documented |
|
Uploaded documents and the results of a processing job have an automatic retention window. A customer can also remove the source file or job artifacts on demand. Those files carry the inputs and outputs for a specific execution. Each new job works from the documents and configuration supplied to it, without a separate store for insights learned across the customer's past jobs. Sourcedocs.reducto.ai/workflows/deletionread 2026-11-20 |
|
| Deployment & Data Residency | Full |
|
Enterprise customers can run Reducto in a dedicated VPC or use a hybrid setup with documents stored in their own cloud account and processing handled by Reducto. Full private installations keep the runtime in the customer's network boundary. Regional cloud processing is also available through endpoints in the European Union and Australia for customers choosing a hosted option. Sourcedocs.reducto.ai/onprem/enterprise_deployment_optionsread 2026-11-20 |
|
| Prebuilt Agents / Templates / Packs | Partial |
|
Studio can start from a sample document and use the buyer's description to build an extraction schema. Reducto supplies cookbook recipes for invoices, contract review and multilingual files. Those starting assets help assemble a workflow. The buyer chooses the processing steps and fields, tests its own files and deploys the resulting pipeline for its application. Sourcereducto.ai/developersread 2026-11-20 |
|
| Triggers & Channel Coverage | Partial |
|
An application submits a document to an asynchronous endpoint and receives a job ID immediately. The job moves through the processing queue, and Reducto can notify a webhook when it completes. Synchronous processing remains available for smaller requests. New work begins with a document submission, while the receiving application's own events or schedules control when it submits. Sourcedocs.reducto.ai/workflows/async-overviewread 2026-11-20 |
|
| Model Flexibility & Routing | Partial |
|
Private installs can set different models for fast and accurate workloads through LiteLLM. The hosted product offers Reducto model variants and version pinning for a supported processing operation. An organization can choose a model routing setup inside its own deployment, while the general cloud service offers choices within Reducto's processing family. Sourcedocs.reducto.ai/onprem/llm_optionsread 2026-11-20 |
|
| APIs / SDKs / MCP Extensibility | Full |
|
Developers can call parsing, extraction, splitting, classification and editing over the API. Studio deployments are accessible through the pipeline endpoint. SDKs support direct integration into Python and Node.js applications and core operations in Go. A command line client and MCP server give scripts and agent clients access to the same document tools. Sourcereducto.ai/developersread 2026-11-20 |
|
| Testing, Debugging & Optimization | Partial |
|
Studio lets a tester run a draft pipeline on sample documents, adjust settings and inspect the resulting data and source highlights before deployment. Smart Schema can help clarify the field definitions used for extraction. This gives a team a cycle for reviewing actual output and tuning a configuration. The sample interface is built around the tester's own inspection of the results. Sourcereducto.ai/studioread 2026-11-20 |
|
| Browser / Computer-use | Not documented |
|
Edit operates directly on supported document files. It uses a field schema or instructions to fill PDF form fields or modify the contents of a Word document and returns the edited file. The processing pipeline handles the document structure and content. It does not drive the Office user interface or navigate through a remote browser session to make those changes. Sourcereducto.ai/developersread 2026-11-20 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Pricing
$150 in free credits, Standard Parse from $1 per 1,000 pages
Operation and pages processed
What is public
Reducto charges $1 per 1,000 pages for Parse on r-1 mini or $10 on r-1 under Standard. Extract and Split are $20, Deep Extract and Deep Split $40, Classify $7.50, Deep Classify $15 and Edit $60 per 1,000 pages, or $15 for fully prefilled pages. Extract, Split and Classify include parsing. Standard also carries up to five Studio seats. Growth and Enterprise rates depend on volume and contract terms.
Billing mechanics
Eligible Parse and Extract jobs run through the batch queue at a 20% discount with a 12 hour completion guarantee. Dense extractions over 100 fields per page can incur a surcharge. Standard includes $150 in free credits for eligible accounts.
Variable cost rationale
Every endpoint bills per page, so spend scales directly with document volume, and each Deep variant costs twice its base rate.
Additional watchouts
Standard terms allow Reducto to use uploaded content and output for training and evaluation. Signed Growth or Enterprise terms cover workloads that require different data handling.
Sales call required
Mixed (some tiers require a call)
Free / trial
$150 in Standard credits for eligible new accounts
Lowest paid plan
Standard
Commercial notes
Teams moving from another document processor can apply for up to $5,000 in migration credits. A startup program offers free credits and discounted Growth pricing to companies under $15 million in funding.
Missing data
Volume rates and contract minimums for Growth and Enterprise are quoted.
Related vendors
- 11th Estate — Agentic platform whose AI engine scans markets, matches an…
- Abacum — FP&A platform for growing companies whose AI Modeler builds, edits…
- Accend — Accend builds human in the loop AI agents for commercial credit…
- Adopt AI — Finance agents that categorize, reconcile and close the books inside…
- Aera Technology — Decision intelligence platform where an always on agent executes…
- Akro AI — On premise operational intelligence that automates document heavy…
Alternatives to Reducto
Other products that do the work Reducto does, and what each one is.
- ChipAgents
- HebbiaAI platform for finance and legal work
- IBM SterlingOrder management and fulfillment platform
- MakinaRocks
- monday.comWork platform
- Box