Most professionals spend hours each week reading contracts they do not fully understand — vendor agreements, NDAs, software licenses, freelance SOWs, employment contracts. The problem is not intelligence; it is that legal language is deliberately opaque and the conventions governing it take years to internalize. Claude handles this translation work efficiently. This guide covers the specific prompts and workflow for using Claude to review legal documents in a fraction of the time, with enough accuracy to catch real issues before they become expensive ones.
One important caveat: Claude is not a lawyer and this workflow does not replace legal counsel for high-stakes contracts. What it does replace is the hours you spend reading a 40-page vendor agreement before knowing whether it is worth your attorney's time at all.
Claude has a large context window — currently 200,000 tokens in Claude 3.5 Sonnet and up to 1 million in some configurations — which means you can paste an entire contract and ask questions about it without the document getting truncated. Most legal documents you will encounter in business are well under this limit. Claude also tends toward careful, qualified language, which is appropriate for legal review: it will flag uncertainty rather than asserting false confidence.
The combination of large context, strong reading comprehension, and calibrated hedging makes Claude the best current AI model for contract review tasks. It consistently outperforms general-purpose prompting approaches on structured extraction and clause-level analysis.
Before diving into clause analysis, get the full picture of what a contract actually commits you to. This prompt creates a clean summary of the core commercial and legal terms:
Here is a contract: [paste full contract text] Extract and summarize the following as a structured list: 1. Parties (full legal names and roles) 2. Contract term and renewal conditions 3. Payment terms (amounts, timing, late fees) 4. Deliverables or services the vendor owes me 5. What I am obligated to provide or do 6. Termination rights (who can terminate, on what notice, for what cause) 7. Liability caps and indemnification obligations 8. Intellectual property ownership 9. Governing law and dispute resolution 10. Any unusual or non-standard provisions For each item, quote the relevant section number alongside your summary.
This output is your contract brief — a one-page summary you can reference throughout the engagement without re-reading the original. It is also what you hand to a colleague or attorney when you need a second opinion.
After extraction, run a targeted risk pass. This prompt focuses on the clauses that most frequently cause problems in practice:
Review the same contract and flag any clauses that: - Create unlimited or uncapped liability for me - Include auto-renewal with short cancellation windows (under 30 days) - Give the other party broad IP rights over work I create - Restrict my ability to work with competitors (non-competes, exclusivity) - Include unilateral modification rights (they can change terms without my consent) - Have liquidated damages or penalty clauses - Contain unusually broad indemnification obligations on my part - Include data sharing or privacy provisions I should be aware of For each flag: quote the exact clause, explain why it is risky, and suggest what a more balanced version would say.
The output from this prompt gives you a specific list of negotiation points rather than vague unease. Most counterparties expect some pushback on standard contracts — knowing exactly which clauses to push back on is the difference between a productive negotiation and being nickel-and-dimed into a bad deal.
When reviewing a revised draft or comparing your standard contract against a modified version, this prompt does the diff work that would normally take an hour:
Here is Version 1 of a contract: [paste V1] Here is Version 2 of the same contract: [paste V2] List every material difference between the two versions: - What was added (quote the new language) - What was removed (quote what was deleted) - What was changed (show before and after) For each change, rate the impact: Favorable to me / Unfavorable to me / Neutral. Explain your reasoning briefly for any Unfavorable changes.
This is particularly useful for software licensing agreements, which frequently contain minor version updates that include material scope changes buried in legal language. Catching a clause that silently expanded the vendor's data rights is worth the 30 seconds it takes to run this prompt.
Once you have identified the clauses you want to push back on, Claude can draft the professional response language. Legal negotiation is largely about tone — you want to be direct without being adversarial, and specific without being inflammatory:
I am negotiating a contract and want to push back on the following clause: [paste the clause] My concern: [explain what bothers you about it in plain language] My preferred outcome: [describe what you want instead] Write: 1. A professional email paragraph explaining my concern and proposed change 2. Alternative contract language that addresses my concern 3. A fallback position if they will not accept my preferred change Keep the tone collaborative, not adversarial. I want to preserve the relationship.
If you regularly review the same types of contracts — vendor agreements, client SOWs, or licensing deals — you can automate the initial review pass entirely. Using Make.com, build a workflow where new contract PDFs dropped into a shared folder trigger automatic text extraction, Claude API review, and a structured summary delivered to your email or project management tool.
The practical result: any new contract gets a first-pass review in minutes rather than hours, and you only spend focused time on the flagged clauses rather than reading every word. For teams that handle high volumes of repetitive contracts — freelancers with standard client agreements, procurement teams reviewing vendor terms — this can reclaim several hours per week. See our guide to using AI for contract review for the full framework including more contract types.
Claude cannot give legal advice, cannot predict how a court would rule on an ambiguous clause, and cannot know jurisdiction-specific regulations that might override contract language. Use Claude for: understanding what a contract says, identifying clauses that warrant a closer look, drafting negotiation language, and comparing versions. Hire an attorney for: high-value transactions, contracts with significant liability exposure, employment agreements with non-competes, equity or IP assignment agreements, and any contract where the stakes exceed the cost of professional review.
The workflow above saves professional time by ensuring that when you do engage an attorney, you are bringing specific questions rather than asking them to read the whole document from scratch. That alone can cut your legal review costs by 40–60%.
💡 Claude is one of the best AI tools for reading dense documents quickly. Browse all recommended AI tools for legal and research workflows →
For contract review and research workflows: Claude directly via claude.ai or the Claude API (no paid tool required for the prompts above). To automate contract ingestion and routing: Make.com handles PDF extraction, Claude API calls, and results delivery without code. For teams doing research-intensive contract work: Notion AI can store and query your contract briefs across all reviewed agreements, building a searchable knowledge base over time. For the broader context of using Claude for research tasks, see our guide on using Claude for research workflows.
Practical prompts and automation ideas — no fluff.