Reducto pricing, free plan and limits
Agentic document platform that parses, extracts and splits PDFs, scans and spreadsheets for AI teams; its About page cites more than 5 billion pages processed.
- Best for
- AI or ML engineer building retrieval over documents
- Not for
- Solo developer who needs a free self-hosted parser
Should you use Reducto?
Reducto suits AI teams that need reliable parsing and extraction from messy documents and want published per-page prices. Its Growth tier adds zero data retention and a HIPAA business associate agreement. G2 shows no reviews yet, so test both parse models on your own files. It has raised $108 million, most recently a $75 million Series B.
Publishes per-page prices for every endpoint and pairs an agentic parse model with extraction, splitting and editing behind one API.
Accuracy and compliance options such as EU-only processing and private deployment create switching costs, while rivals such as LlamaParse sell parsing alone and cloud OCR services bundle it with their platforms.
Also for
- Platform engineer at a regulated company that needs private deployment
- Agent builder wiring document reading into Claude Code or Cursor
Show 4 more
- AI and ML engineers building retrieval pipelines that ingest PDFs, scans and spreadsheets
- Platform engineers at regulated fintech, legal and healthcare companies that need EU-only, VPC or on-premise document processing
- Data teams replacing a cloud OCR service such as AWS Textract, Azure Document Intelligence or Google Document AI
- Agent builders who want Claude Code, Cursor or Codex to read documents through MCP
Also not for
- Non-technical operator who wants a finished workflow app
Main trade-off
Growth and Enterprise pricing is quote-only, and its accuracy claims rest on its own evaluations.
Reducto pricing
| Plan | Monthly price | What it includes |
|---|---|---|
| Standard | Free | Pay as you go at published per-page rates, with $150 in free usage, Parse, Extract, Split and Edit APIs, 30+ supported file types, Up to 5 seats in Reducto Studio, No page limits |
| Growth | Custom | Everything in Standard, Volume discounts, Zero data retention agreement and a Business Associate Agreement, Premium rate limits and priority requests, Priority Slack, Teams and email support, EU and AU data residency endpoints |
| Enterprise | Custom | Everything in Growth, VPC and on-prem deployments, Custom MSA and SLA, Custom rate limits, throughput and processing pipelines, Dedicated on-call support, Role-based access control, SSO and SAML |
How the pricing works, in full
List prices took effect on 1 September 2026, per Reducto's docs: Parse with r-1 is $10 per 1,000 pages and r-1 mini $1, Extract $20, Deep Extract $40, Split $20, Deep Split $40, Classify $7.50, and Edit $60 ($15 when pages are prefilled). Parse and Extract batch jobs get 20% off. The Standard plan includes $150 of free usage and up to 5 Studio seats. Growth and Enterprise are custom-priced after a sales conversation, and Reducto offers up to $5,000 in migration credits to teams switching from another processor. Contracted rates can differ from list. The About page still mentions 15,000 free credits; this page follows the pricing page and docs. Legacy Parse costs $15 per 1,000 pages for a standard page.
Before you commit
- Prompted blocks cost $5 per 1,000 blocks and the advanced chart agent $0.06 per chart, on top of the page rate
- r-1 mini needs a $1 per 1,000 pages add-on for OCR data return or Parse citations
- Dense extractions over 100 fields per page may incur a per-field surcharge
What Reducto does
Reducto is an agentic document platform that parses, extracts, splits, classifies and edits PDFs, scans and spreadsheets into structured, citation-ready data through an API, a no-code Studio and MCP servers. Its r-1 parse model became generally available in November 2026, with a cheaper r-1 mini beside it. Choose it for high-volume AI document pipelines.
Parse with r-1 and r-1 mini
r-1 is the flagship parse model and r-1 mini a cheaper one; both include agentic capabilities for complex documents, per the November 2026 changelog.
Extract and Deep Extract
Extract v4 is now the default extraction model, with citations and confidence scores, and Deep Extract is the higher-priced option, which a July 2026 Reducto post pitches for near-perfect accuracy on hard files.
Split, Deep Split and Classify
Splits multi-document files into units, routes documents by type before processing, and offers an agentic Deep Split for long files with many categories.
Edit API
Fills and edits documents programmatically, at a lower rate when pages are fully prefilled.
Show all 8 features
Studio and Smart Schema
Reducto Studio builds and deploys document pipelines without code, and Smart Schema drafts and improves extraction schemas automatically.
Local and hosted MCP servers
A uvx server for local files and a hosted server for public URLs let agents classify, parse, extract, split and edit documents.
Batch Queue
Parse and Extract jobs sent through the queue are discounted and carry a completion guarantee, which suits large asynchronous backlogs.
Deployment options
Offers hosted cloud, EU-only processing, hybrid VPC and on-premise installs, with VPC and on-prem listed under Enterprise.
Works with
- Claude Desktop
- Claude Code
- Codex
- Cursor
- VS Code
- Windsurf
- Python and Node SDKs
- AWS Marketplace
Capabilities
- Document parsing
- OCR
- Data extraction
- Agentic OCR
- MCP
- RAG infra
Strengths and limits
Strengths
- Every endpoint has a published per-page price, so a team can model cost before signing anything.
- Growth and above get a zero data retention agreement, under which API-submitted data expires within 24 hours, and the data policy rules out training on document content.
- Local and hosted MCP servers cover Claude Desktop, Claude Code, Codex, Cursor, VS Code and Windsurf, with a documented install command for each.
- SOC 2 Type 2 and a HIPAA-compliant pipeline with a signed BAA are listed for Growth and Enterprise customers in the docs.
- The customers page names Harvey, Scale AI, Vanta, Legora and Rogo, and an independent Built In article confirms Harvey, Rogo and Scale AI as users.
Limits
- Pricing moved from credits to per-endpoint page rates on 1 September 2026, so older estimates no longer apply, and the About page still cites the old 15,000 free credits.
- Growth and Enterprise are quote-only, and VPC and on-premise deployment sit in Enterprise, so regulated buyers cannot see their rate before a sales call.
- G2's page for Reducto says it has not been reviewed yet, so independent user feedback is missing.
- The accuracy results it publishes come from its own ParseArena and PAIR evaluations, and r-1 mini does not process spreadsheets, per the docs.
- Dense extractions over 100 fields per page can add a per-field surcharge, and Deep Extract is not eligible for the batch discount.
Reducto alternatives
| Tool | Price from | Free plan | Runs on | Rating |
|---|---|---|---|---|
| Reducto This page | -- | Yes | Web | 3.9 |
 ZeroxPick Zerox if a free MIT-licensed library appeals and you will pay your own vision-model bill; pick Reducto for a managed API with extraction, splitting and a zero-retention option. | Free | Yes | -- | 3.2 |
 LlamaIndexPick LlamaIndex if you already build on its framework and want LlamaParse credits on a $0 plan; pick Reducto if per-page pricing and a dedicated extraction and editing API matter more. | Free | Yes | Web | -- |
 ExaSearch API that ranks the web by meaning for AI agents and retrieval pipelines, with Lite, Auto and Deep search modes, an asynchronous research Agent and an official MCP server. A free Starter plan with monthly credits needs no payment method. | $49/mo | Yes | Web | 4.0 |
Reducto facts
- Training-data policy
- A zero data retention policy applies from the Growth tier up: submitted API data is deleted after a day, other than Studio jobs and results kept with persist_results. The docs add that customer document content is not used for model training, and they state no policy for the Standard plan (Reducto data policies docs, read 12 November 2026).
- Compliance
- SOC 2 Type II · HIPAA (BAA on Growth and Enterprise) · EU-only processing option
- Works offline
- Yes
- Minimum skill level
- Intermediate
- Price band
- Mid-range
- Total funding
- $108M: $8.4M led by First Round Capital (Oct 2024), $24.5M Series A led by Benchmark (Apr 2025) and $75M Series B led by a16z (Oct 2025)
Data and privacy
Proprietary
Questions about Reducto
How much does Reducto cost in 2026?
Reducto's list prices from 1 September 2026 are $10 per 1,000 pages for Parse with r-1, $1 for r-1 mini, $20 for Extract, $40 for Deep Extract, $20 for Split and $60 for Edit, with a batch discount on Parse and Extract. Growth and Enterprise are custom-priced, and contracted rates can differ from list. Dense extractions and prompted blocks can add small charges.
Is Reducto free to use?
Partly. The Standard plan starts with $150 in free usage, up to 5 Studio seats and no page limits, and signup is open to personal email addresses. After that you pay the per-page rates. Zero data retention, a BAA and EU or AU endpoints need the custom-priced Growth plan.
Does Reducto keep or train on my documents?
For Growth and above, API-submitted data expires within 24 hours, except Studio jobs and results saved with persist_results, and Reducto says it does not train models on document content. Its docs state no retention or training policy for the Standard plan. Enterprise adds VPC and on-prem deployment, and EU-only processing is documented.
How do I connect Reducto to Claude Code or Cursor?
Reducto's docs show a local server and a hosted one. For Claude Code, run uvx mcp-server-reducto --login, then claude mcp add -s user reducto -- uvx mcp-server-reducto. The hosted server at mcp.reducto.ai uses an API key in an Authorization header. Cursor, Codex, VS Code and Windsurf have their own snippets.
How does Reducto compare with LlamaParse and Zerox?
LlamaIndex prices LlamaParse in credits, with a $0 plan that includes 10K free credits, so convert credits to pages for your parse mode before comparing with Reducto's per-page rates. Zerox is a free library under the MIT license that passes page images to vision models, and its latest commit is more than a year old. Reducto sells a managed API with extraction, splitting and editing.
About Reducto
Reducto is a San Francisco-based document platform. It introduced its Document API on 27 February 2024, and Built In's October 2025 coverage names CEO Adit Abraham and CTO Raunak Chowdhuri as co-founders. It has raised $108 million in total: $8.4 million led by First Round Capital in October 2024, a $24.5 million Series A led by Benchmark in April 2025, and a $75 million Series B led by a16z on 14 October 2025 alongside Benchmark, First Round Capital, Y Combinator and BoxGroup. On 7 May 2026 it bought Opennote, an AI notebook for students.
Read the full background
Its pages call it an agentic document platform: Parse, Extract, Split, Classify and Edit APIs, Reducto Studio for building pipelines without code, a CLI and an MCP server. The parse model changed in 2026. r-1 entered preview in August and became generally available in November, and r-1 mini launched on 5 November at a tenth of the price of Parse r-1. Reducto's own ParseArena test, judged by its open PAIR framework, says r-1 mini won more pairwise comparisons than it lost against Google's Gemini Layout Parser at 90% lower cost, and that Reducto r-1 beat every parser it tested on win rate. Those are vendor-run results, and the post itself tells buyers to test on their own documents. From 7 December 2026 requests that set no model will use r-1.
Extraction moved too. Extract v4 became the default in September 2026 with better citations and confidence scores, and Deep Extract, introduced on 6 April, adds an agentic pass for long, high-stakes documents. Classify arrived on 21 May, Deep Split on 3 June and Smart Schema in Studio on 8 June, and a Batch Queue discounts Parse and Extract jobs with a 12-hour completion guarantee. Reducto has been on AWS Marketplace since 6 February 2026. Agents reach it through a local uvx MCP server or a hosted one at mcp.reducto.ai, which the docs say work with Claude Desktop and Code, Codex, Cursor, VS Code and Windsurf.
Its customers page lists Harvey, Scale AI, Vanta, Legora, Rogo, JLL and Datasite among others. Harvey and Rogo are both on Hokai, and the About page says the platform has processed more than 5 billion pages. It suits developers and platform teams building document features for retrieval and agents; see AI search and retrieval infrastructure for the wider shelf.
Who should look elsewhere: teams that want free, self-hosted parsing or that process simple documents at the lowest per-page cost. Zerox is an MIT-licensed library that sends page images to vision models, and its repository shows 12.3k stars but no commit since May 2025. LlamaIndex sells LlamaParse from a $0 plan with free credits. Reducto's own benchmark post also tests parsers from Azure, Amazon Textract, Mistral, Datalab and Extend, so a buyer comparing vendors has a vendor-run starting list but no independent test.
Sources and guides
Official sources
- Reducto official site Vendor
- Reducto documentation Docs
- Reducto MCP server guide Docs
- Reducto usage and pricing Docs
- Reducto blog Docs
- GitHub repository GitHub
Product facts on this page (pricing, plans, integrations, compliance) are read from Reducto's own pages and documentation. The verdict is Hokai's reading of them, not the vendor's claim.
Hokai guides covering Reducto
Also mentioned in
Lists that include Reducto
More AI Tools on Hokai
LanceDBOpen-source embedded retrieval library with a distributed Enterprise lakehouse on the same Lance format; its GitHub repository shows 11.9k stars.
Mem0Memory layer for AI agents: an Apache-2.0 SDK, a managed Platform and an MCP server, with a free plan and 68.8k GitHub stars.- AlgoliaAI search and retrieval platform powering 1.75 trillion searches annually for 18,000+ businesses with semantic search, vector embeddings, and AI agents.
DeeplakeActiveloop's open-source GPU database for AI agents: Postgres interface, multimodal data lake, SOC 2 Type II certified, Apache-licensed core.
BrowserbaseHosted headless browsers for AI agents, with a free plan of 3 concurrent sessions, session replays and the open source Stagehand SDK.FirecrawlWeb data API for AI agents: search, scrape, crawl and interact, returning LLM-ready Markdown. Open source core with a free plan and an MCP server.