---
title: "Legal AI agents · contract review, due diligence, legal research · Agent Fieldbook"
url: https://agentfieldbook.org/catalog/legal/
description: "Legal agents built for a domain where precision and provenance beat fluency: contract review and drafting, due-diligence analysis, legal research and discovery. Retrieval-grounded and human-in-the-loop by necessity, each specced against primary sources."
section: "Catalogue · Legal"
source: Agent Fieldbook — generated from the published page
---

# Legal agents

Agents built for legal work: contract review and drafting, due-diligence analysis, legal research and discovery. A domain where precision and provenance beat fluency, because a confident wrong answer carries liability. The credible ones ground every claim in a source and stay human-in-the-loop. Click a card for its full spec.

## 3 agents Live data · click a specimen for its anatomy

## Field notes

Precision and provenance, where a confident wrong answer carries liability

Legal agents cluster around a narrow set of tasks: contract review, discovery, research and drafting. All four are document-heavy, high-volume and reviewed by a professional who carries personal liability for the output.

That liability shapes the products. Review is not a feature bolted on for comfort, it is the operating model, and citation to source is mandatory rather than nice to have. The agents that succeed here are the ones that make verification fast, not the ones that ask to be trusted.

- [Brandon Chaplin](https://www.linkedin.com/in/brandon-chaplin-digital-marketing-strategist)

## Common questions

### What can AI agents do in legal work?

Mostly reading at volume: contract review and comparison, due-diligence document analysis, legal research and first-draft clauses. The tasks that pay off are the ones where someone would otherwise read hundreds of pages to find a handful of issues.

### Do AI legal tools make up cases?

They can. Language models generate citations that look right but do not exist, and lawyers have been sanctioned for filing them unchecked. Tools that retrieve from a real case database and link every citation cut the risk, but each citation still needs opening and reading.

### Can I use ChatGPT for contract review?

For a first pass, yes. For anything you rely on, not without review. A general assistant has no access to your precedent library or your negotiation playbook, and client confidentiality limits what you can paste into it. Those gaps are why purpose-built legal tools exist.

### Will AI replace lawyers?

Not in the deployments documented so far. It removes hours of reading and searching, but the advice, the judgement and the professional liability stay with the lawyer. Someone regulated still has to sign the work.

## Next in the learning path

- [Vertical SaaS agents The broader pattern legal sits inside](https://agentfieldbook.org/catalog/vertical_saas/)

- [Finance agents Another precision-and-provenance vertical](https://agentfieldbook.org/catalog/finance/)

- [Retrieval patterns How legal agents ground their answers](https://agentfieldbook.org/memory/retrieval/)

- [Human oversight Why legal work stays human-in-the-loop](https://agentfieldbook.org/security/oversight/)

## Entries

_3 entries listed on this page._

| name | description | subjectOf |
| --- | --- | --- |
| Harvey AI | Market-leading enterprise legal agent, expanded 6→60+ jurisdictions with autonomous agents; table stakes at $50M+/yr firms. | https://www.harvey.ai |
| Legora | Collaborative legal AI calling 2026 "the year of agents in legal": agents completing end-to-end legal work with human oversight. | https://news.crunchbase.com/venture/unicorn-legal-tech-ai-startup-legora-triples-valuation/ |
| Spellbook | Word-native contract drafting and review agents for transactional lawyers. | https://www.businesswire.com/news/home/20251009110230/en/ |