Contracts AI that makes your LLM pipeline better
You could send every page of every contract through an LLM. Or you could use purpose-built ML to find the relevant clauses first, then pass structured results to the LLM for interpretation. Faster. Cheaper. More reliable.
What Zuva API is
A REST API for extracting structured data from contracts. 1,400+ pre-built fields for legal provisions. Document classification. Custom field training. Works as a standalone tool or as a layer in your AI pipeline.
See capabilitiesTry it free
Sign up, get an API token, and start making requests. Free tier includes 25 files, 25 extractions, and 25 classifications per day. No credit card required.
Create free accountDocumentation
Quickstart guides, authentication, endpoints, and code examples. Plus the full API reference with request/response schemas.
Documentation API ReferenceWhy not just send everything through an LLM?
You could. But for large-scale contract work, it's expensive, slow, and less reliable than you'd expect. LLMs are powerful interpreters, but they're not great at systematically finding specific provisions across thousands of pages. That's what purpose-built ML does well.
Sending full contracts to an LLM
Zuva ML → then LLM
How Zuva uses our AI + an LLM to generate better results than either alone: Our "Answers" feature uses Zuva ML to find relevant clause text (e.g., change-of-control provisions), then passes those extractions to an LLM to select a structured answer (e.g., "consent required" / "notice required" / "freely allowed"). The ML finds. The LLM interprets. The combination is more reliable than either alone.
What you can do with the API
Extract contract provisions
Pull specific clauses from contracts using 1,400+ pre-built fields covering change of control, assignment, termination, IP, non-competes, and more. Each field was trained using tens of thousands of hours of work by experienced lawyers, many from firms like Goodmans, Sidley, Allen & Overy, Stikeman, and Torys.
Classify documents
Automatically identify 225+ document types. Know whether you're looking at an NDA, lease, employment agreement, or amendment before running extraction.
Detect language
Identify the predominant language of a document. Useful for routing in multi-language contract portfolios.
Self-host or choose your region
Self-host on Azure or AWS out of the box, or work with us to deploy in other environments for large enterprise contracts. Most customers use our cloud-hosted API on Azure in the region of their choice (typically US or EU). You control data retention: no retention, a week, or whatever policy fits your requirements.
Feed your pipeline
Use Zuva as the extraction layer in a larger AI pipeline. Pass structured clause data to your own LLM, RAG system, or application logic. Connects directly to Power Automate via our connector, or integrate via REST API with any platform.
What people build with the API
Contract lifecycle management (CLM)
Populate your CLM with structured data from existing contracts. Auto-extract key terms on upload.
Due diligence automation
Build M&A diligence workflows that extract deal-relevant provisions and flag risk automatically.
Virtual data rooms
Add AI-powered contract intelligence to your VDR product. Classify documents and extract provisions on ingestion.
Compliance and audit tools
Scan contract portfolios for specific provisions in response to regulatory changes or policy updates.
Zuva has a Power Automate connector
If your team uses Microsoft's Power Platform, you can connect Zuva directly into your workflows without writing code. Build contract review automations that run whenever a document hits a SharePoint folder, an email arrives, or a form is submitted.
Microsoft's own legal team built a workflow using Zuva's API in Power Automate to auto-triage low-risk contract reviews. Their proof of concept was working within about an hour. The result: roughly 75% reduction in time spent on routine contract requests, with a target of going from a 3-day SLA to about 3 minutes.
Two lawyers from Microsoft Legal describe how they built a contract review automation using Zuva's API and Power Automate.
What people are saying
Start free. Scale as you need.
Free
Test the API with daily limits. No credit card required.
Pay-as-you-go
$5,000 minimum commitment. Bulk discounts and flat-fee arrangements available for larger volumes.
Enterprise
For teams embedding Zuva into a product or running large-scale ongoing analysis. Multi-product discounts available.
Start building
Documentation
Quickstart guides, authentication, endpoints, and code examples to get up and running.
View docsAPI Reference
Full endpoint reference with request/response schemas and field IDs.
View referenceAI Fields Library
Browse all 1,400+ pre-built extraction fields by category, document type, and jurisdiction.
Browse fieldsCommon questions
Why would I use Zuva instead of just prompting an LLM?
LLMs are great at interpreting text, but they're not optimized for systematically finding specific provisions across large document sets. They're expensive per page, inconsistent across runs, and biased toward plausible-sounding answers rather than exhaustive recall. Zuva's ML is purpose-built for clause extraction: it's consistent, high-recall, and far more cost-efficient at scale. The ideal approach is to use Zuva to find the relevant material, then pass it to an LLM for interpretation.
Can I self-host the API?
Yes. Our models are data-efficient and can self-host on Azure or AWS out of the box. For large enterprise engagements, we can work with you to deploy in other environments. Most customers use our cloud-hosted API on Azure, choosing their preferred region (typically US or EU). Either way, you control data retention policy.
What's the minimum commitment?
The pay-as-you-go plan has a $5,000 minimum commitment. The free tier has no commitment and resets daily, so you can test the API thoroughly before deciding. Bulk discounts and flat-fee arrangements are available for larger volumes.
How is this different from RAG?
RAG (retrieval-augmented generation) typically uses topic-similarity search to find possibly-relevant chunks of text, then passes those to an LLM. The problem with contracts is that generic similarity search often misses legally significant provisions or returns irrelevant text. Zuva's extraction is trained specifically on legal provisions. It knows what a change-of-control clause looks like, even when phrased in unusual ways. It's a more accurate retrieval layer for the contract domain than generic vector search.
Where is my data hosted and how is it secured?
Zuva is SOC 2 Type II certified. Our cloud-hosted API runs on Azure, and you choose the region (most customers pick US or EU). You set your own data retention policy: documents can be deleted immediately after processing, retained for a set period, or anything in between. For self-hosted deployments, Azure and AWS are supported out of the box. For large enterprise contracts, we can work to support other environments.