Unstract
Los Altos, California, United States
What it is
Unstract extracts structured data from unstructured documents and automates document workflows. Prompt Studio uses natural-language instructions to define extraction projects and deploy them as APIs or ETL pipelines. Agentic Prompt Studio uses multi-agent pipelines to set schemas, create prompts, and validate extraction across documents. The platform describes its processing as document agnostic, without a need for prior training or templates. Human-review workflows can route selected documents for approval and validation. LLMChallenge uses two language models to challenge extracted responses; SinglePass and Summarized Extraction are described as ways to reduce token use. Integrations include Anthropic, OpenAI, and Vertex AI models, along with file systems and databases such as Azure Cloud Storage, PostgreSQL, and S3/Minio. Deployment options are managed cloud, on-premise enterprise, and open-source self-hosted. The free Open Source plan is self-hosted and limited to one user. Paid Starter begins at $499 per month, with 5,000 pages monthly and $0.10 per overage page. Unstract says cloud data is encrypted in transit and at rest and is not used for training.
For and against
For
- Deploys extraction projects as APIs or ETL pipelines.
- Supports human review and validation.
- Connects to multiple model providers and data sources.
- Offers managed cloud, on-premise, and self-hosted editions.
- Cloud data is encrypted and not used for training.
Against
- Free Open Source plan is limited to one user.
- Starter monthly plan includes 5,000 pages.
- Starter overage costs $0.10 per page.
- The free edition is self-hosted.
The verdict
Unstract combines document extraction with workflow deployment and optional human review. Its free edition has a one-user limit, while paid page allowances and overage charges should factor into plan selection.
Who it suits
Unstract may suit teams that need to extract data from varied documents and route selected items for human review. Its self-hosted, managed-cloud, and on-premise options address different deployment preferences.
Plans and prices
LLMWhisperer Native Text
digital PDFs · $0.001/page · 100 free pages/day
Read 15 Oct 2026 · unstract.com
LLMWhisperer Low Cost
clean scans · $0.005/page · 100 free pages/day
Read 15 Oct 2026 · unstract.com
LLMWhisperer High Quality
hard scans/handwriting · $0.01/page · 100 free pages/day
Read 15 Oct 2026 · unstract.com
Document Studio Starter annual
60,000 pages/year · $0.10/page overage · Agentic Prompt Studio · standard compliance/support
Read 15 Oct 2026 · unstract.com
Document Studio Starter monthly
5,000 pages/month · $0.10/page overage · Agentic Prompt Studio · standard compliance/support
Read 15 Oct 2026 · unstract.com
Document Studio Growth annual
300,000 pages/year · $0.09/page overage · Agentic Prompt Studio · standard compliance/support
Read 15 Oct 2026 · unstract.com
Document Studio Growth monthly
25,000 pages/month · $0.09/page overage · Agentic Prompt Studio · standard compliance/support
Read 15 Oct 2026 · unstract.com
Open Source
1 user only · self-hosted · Slack community support
Read 4 Oct 2026 · unstract.com
Starter (annual)
60,000 pages/year · $0.10/page overage · standard compliance and support
Read 4 Oct 2026 · unstract.com
On the record
- Agentic extraction
- Agentic Prompt Studio uses multi-agent pipelines to define schemas, generate extraction prompts, and validate accuracy across documents.unstract.com · read 4 Oct 2026
- Agentic Prompt Studio
- Agentic Prompt Studio can generate schemas, craft prompts and validate extraction accuracy.unstract.com · read 15 Oct 2026
- API data retention
- In asynchronous extraction, text is held in server memory for 60 minutes; enabling highlighting stores text and metadata on the servers.docs.unstract.com · read 15 Oct 2026
- API modes
- The extraction API supports OCR or text processing, line-printer or text output, page selection and synchronous or asynchronous responses.docs.unstract.com · read 15 Oct 2026
- Changelog news
- The August 12, 2026 cloud release added hosted MCP endpoints beside API deployments and global API deployment keys.docs.unstract.com · read 15 Oct 2026
- Company size and headquarters
- LinkedIn lists Unstract as a privately held 11–50 employee company headquartered in Los Altos, California.linkedin.com · read 15 Oct 2026
- Connector examples
- The connector documentation gives Azure Cloud Storage, PostgreSQL, and S3/Minio as examples.docs.unstract.com · read 4 Oct 2026
- Deployment
- The product is offered as managed cloud, on-premise enterprise, or open-source self-hosted editions.unstract.com · read 4 Oct 2026
- Deployment options
- Workflows can be deployed as REST APIs, ETL pipelines or MCP servers, with cloud and self-hosted editions.unstract.com · read 15 Oct 2026
- Document agnostic
- The maker says Unstract handles documents of any format without prior training or templates.unstract.com · read 15 Oct 2026
- Documentation/support
- Official documentation covers editions, quick start, MCP, CLI, Prompt Studio, workflows, API deployments, ETL pipelines, connectors, adapters and a platform API.docs.unstract.com · read 15 Oct 2026
- Document support
- Unstract describes itself as document agnostic and says it works with documents without prior training or templates.unstract.com · read 4 Oct 2026
- Enterprise controls
- The feature comparison lists SOC 2 Type II, ISO 27001, HIPAA, GDPR with DPA, SSO/SAML, and audit logs for enterprise editions.unstract.com · read 4 Oct 2026
- Extraction checks
- LLMChallenge uses two LLMs to challenge extracted responses, and the product describes SinglePass and Summarized Extraction as ways to reduce token use.unstract.com · read 4 Oct 2026
- Funding and investors
- Crunchbase lists one seed round and Lightspeed India Partners as an investor; LinkedIn also says the company is backed by Together Fund.crunchbase.com · read 15 Oct 2026
- Human review
- The platform provides side-by-side source-document comparison, highlighting and human approval workflows.docs.unstract.com · read 15 Oct 2026
- Integrations
- Listed connectors include S3, Google Drive, Dropbox, Box, Snowflake, BigQuery, PostgreSQL, SQL Server, QuickBooks, Xero, SAP, NetSuite, Salesforce, HubSpot, Jira, Asana, ClickUp and Linear.unstract.com · read 15 Oct 2026
- Legal name
- Crunchbase identifies the legal entity as Zipstack, Inc.crunchbase.com · read 15 Oct 2026
