AI Analysis

AI Analysis Pipeline

Clausemaster uses a three-step AI pipeline powered by the Vercel AI SDK to analyze contracts. Choose between our free community model or bring your own API key for premium providers. Each step builds on the previous one to provide comprehensive results.

Pipeline Overview

1

Classification

Detect contract type & metadata

2

Extraction

Break down into clauses

3

Issue Flagging

Flag risks & concerns

All three steps run sequentially using generateObject from the Vercel AI SDK, producing structured JSON output validated against typed schemas.

Step 1: Classification

The first step identifies what kind of contract this is and extracts key metadata.

Detected Fields

Contract Type

e.g., NDA, Employment Agreement, SaaS Terms, Lease

Jurisdiction

Governing law and jurisdiction identified from the text

Parties

Names and roles of all parties to the agreement

Effective Date

Start date or execution date if mentioned

Step 2: Clause Extraction

The contract is broken down into individual clauses, each analyzed for content, category, and bias.

Per-Clause Analysis

Clause Type

Specific type (e.g., Termination, Indemnification, Non-Compete)

Category

Broad category grouping related clause types

Summary

One-line plain-language summary of the clause

Full Text

The exact text extracted from the contract

Bias Assessment

Neutral, Favors Party A, or Favors Party B

Skill Match

Confidence score if matched to a known legal skill

Bias Assessment

NeutralFavors Party AFavors Party B

Step 3: Issue Flagging

The final step scans for potential problems, missing clauses, unusual terms, and jurisdiction-specific concerns.

Issue Types

Missing Clause

A standard clause expected for this contract type is absent.

Unusual Term

A clause contains non-standard or potentially problematic language.

Jurisdiction Risk

Terms that may conflict with the identified governing jurisdiction.

Ambiguous Language

Vague wording that could lead to differing interpretations.

Compliance Gap

Potential regulatory or compliance issue detected.

Imbalanced Term

A clause significantly favors one party over the other.

Severity Levels

High— Requires immediate attention
Medium— Should be reviewed
Low— Informational concern

Supported Contract Types

The AI can analyze a wide range of contract types. Common examples include:

NDA / ConfidentialityEmployment AgreementSaaS / Software LicenseLease / RentalService AgreementPartnership AgreementTerm SheetPurchase AgreementConsulting AgreementFreelance / ContractorLoan AgreementSettlement Agreement

AI Providers

Clausemaster supports multiple AI providers. Choose the option that best fits your needs and budget.

Community (Free)

Start analyzing contracts immediately with no API key required. Powered by an open-weights model hosted by the platform.

  • No API key or account needed
  • Good for getting started and drafting skills
  • All pipeline features available

Bring Your Own Key

Use your preferred AI provider for higher quality output or specific model requirements.

  • Anthropic (Claude)
  • OpenAI (GPT)
  • Groq (Llama)
  • Mistral
  • Together AI (Llama)

How It Works

Your AI provider choice is stored per-account. You can switch providers at any time from Settings. All providers use the same analysis pipeline and produce the same structured output format — only the underlying model differs. API keys are encrypted at rest using AES-256-GCM.

LegalSkills Integration

The analysis pipeline references the LegalSkills knowledge base during clause extraction and issue flagging to improve accuracy.

How It Works

The LegalSkills repository contains structured legal knowledge (clause patterns, boilerplate language, and best practices). During Steps 2 and 3, this knowledge is loaded as read-only reference context for the AI, improving clause identification accuracy and enabling skill match confidence scores.