Claude Computer Use Review: Hands-On Testing (2026)
When Anthropic released a legal plugin for Claude Cowork on January 30, 2026, RELX dropped 14%. Wolters Kluwer fell 10.5%. Thomson Reuters shed 14.2% in a single trading session. European publishers and legal data companies collectively lost billions in market value.
That’s an outsized reaction for what amounts to an open-source plugin for contract review.
So what’s actually going on here? I’ve spent the past week digging into Anthropic’s legal plugin, comparing it to established tools, and trying to understand why Wall Street panicked while legal tech insiders mostly shrugged.
Quick Verdict
Aspect Rating Overall Score ★★★☆☆ (3.5/5) Best For In-house legal teams with existing Claude subscriptions Pricing Free (requires Claude Cowork subscription) Contract Review Good for basic playbook-based review Enterprise Readiness Limited (no proprietary data, basic integrations) Customization Excellent (open-source) Bottom line: A solid starting point for legal AI, not a Harvey killer. Best suited for in-house teams doing high-volume, routine contract work.
Let’s cut through the hype. Anthropic’s legal plugin for Claude Cowork provides five core capabilities:
Feed it a contract and a negotiation playbook, and Claude runs a clause-by-clause analysis. You get green/yellow/red flags for risk levels plus suggested redlines. This is useful but not revolutionary. Ironclad and Kira Systems have done this for years with deeper customization.
Incoming NDAs get automatically sorted into three buckets: standard approval (sign it), counsel review (quick look needed), or full review (problems detected). For companies processing dozens of NDAs monthly, this could save significant attorney time on routine screening.
Monitors vendor agreement status, flagging upcoming renewals and compliance issues. Basic contract lifecycle management without the full CLM platform.
Generates contextual briefings for daily updates, research summaries, or incident response. Think of it as Claude’s extended thinking applied to legal context.
Templates responses for common requests: data subject access requests, discovery holds, standard inquiry responses. Useful for the repetitive communications that eat in-house attorney time.
The market reaction reveals more about investor anxiety than Anthropic’s actual competitive position.
Here’s what spooked investors: Anthropic bundled model + wrapper + workflow into a single offering. Most legal tech companies are essentially wrappers around foundation models like Claude or GPT-4. If Anthropic (or OpenAI) starts building those wrappers themselves, what’s left for the middleware vendors?
Here’s what investors missed: Proprietary data is the moat. Westlaw and LexisNexis aren’t valuable because of their interfaces. They’re valuable because of decades of curated, verified legal content. Anthropic’s plugin has zero proprietary legal data. It’s Claude’s general knowledge applied to legal documents.
The plugin also lacks the organizational-scale features enterprise legal departments require: centralized policy management, matter-level analytics, compliance audit trails. As Jenn McCarron (co-founder of Contracts.ai) noted, there’s a meaningful difference between individual productivity tools and enterprise platforms.
In-house legal teams at companies already paying for Claude: If you’re on Claude Team or Enterprise, this is a free addition to your existing subscription. For high-volume NDA screening or routine contract review, the ROI is immediate.
Legal ops professionals comfortable with customization: The plugin is open-source. If you can modify Python scripts, you can adapt the playbook logic to your company’s specific negotiation positions. This flexibility doesn’t exist with most commercial CLM tools.
Small legal departments without existing contract software: If you’re choosing between spreadsheets and this plugin, choose this. It’s not a full CLM platform, but it’s significantly better than nothing.
Firms needing verified legal research: Claude will confidently cite cases that don’t exist. For anything requiring legal authority, you still need Westlaw or Lexis+ with their hallucination-resistant grounding.
Large law firms doing complex M&A due diligence: Kira Systems can process thousands of contracts with pre-trained intelligence about hundreds of clause types. Anthropic’s plugin requires you to define your own playbooks. For sophisticated deal work, Kira remains the standard.
Organizations requiring matter-level analytics: Enterprise legal departments need to track performance across matters, attorneys, and contract types. This plugin provides document-level analysis without the portfolio-level insights tools like Evisort deliver.
Everyone wants to know: does this threaten Harvey?
Harvey’s leadership publicly stated the announcement doesn’t change their strategy. I believe them.
Harvey has spent years building legal-specific training, law firm integration infrastructure, and enterprise security certifications. They’ve signed deals with AmLaw 100 firms that require on-premise deployment options, matter-specific fine-tuning, and privileged communication handling.
Anthropic’s plugin is a good starting point that demonstrates Claude’s legal capabilities. Harvey is a production-grade platform built for law firm economics and risk profiles. Different products for different buyers.
The real competition isn’t Anthropic vs. Harvey. It’s whether commodity AI capabilities will squeeze the margins of every legal tech vendor who built thin wrappers over foundation models. That’s a legitimate concern, but it predates this specific plugin.
If you want to try it:
The setup isn’t plug-and-play. You’ll need someone comfortable with MCP integrations and legal workflow design. For most companies, that means involving both IT and legal ops.
Open-source approach: Legal departments can customize without waiting for vendor roadmaps. If your company negotiates specific provisions uniquely, you can encode that logic yourself.
MCP integration: Rather than forcing a proprietary ecosystem, Anthropic built on their open standard. Your legal plugin can connect to the same infrastructure as your other Claude Cowork workflows.
Honest positioning: Anthropic explicitly states all outputs should be reviewed by licensed attorneys. They’re not pretending AI can replace legal judgment. That’s refreshing compared to vendors who overpromise.
No proprietary legal intelligence: Claude’s training includes legal content, but it’s not the curated, verified data that Westlaw and Lexis spend billions maintaining. For research-dependent work, this plugin doesn’t compete.
Basic analytics: You can review contracts, but you can’t easily track trends across your portfolio. What percentage of vendor contracts include problematic indemnification language? The plugin doesn’t tell you.
Limited audit trail: For regulated industries needing to demonstrate compliance processes, the plugin’s logging capabilities are underdeveloped compared to purpose-built CLM platforms.
Anthropic’s legal plugin for Claude Cowork represents something significant: a foundation model company shipping domain-specific tooling rather than leaving it to third parties. That shift in strategy matters more than the plugin’s current capabilities.
For in-house legal teams already using Claude, the plugin offers genuine utility at no additional cost. NDA triage alone could justify the time to set it up.
But this isn’t the legal tech apocalypse some feared. Companies processing complex transactions, needing verified research, or requiring enterprise-scale analytics should continue using specialized tools. Harvey, Kira, and the established legal tech stack aren’t obsolete.
The stock market panic tells you more about investor fear of AI disruption than about Anthropic’s actual competitive position. Legal tech has genuine moats in proprietary data and enterprise relationships. This plugin doesn’t breach them.
What it does do: demonstrate that Claude can handle legal workflows competently, with enough customization flexibility that motivated legal ops teams can build something useful. That’s worth attention, even if it’s not worth a 14% stock decline.
Verdict: A meaningful addition to Claude’s capabilities, not a category killer. Best suited for in-house teams doing high-volume routine work who already have Claude subscriptions.
The plugin itself is free and open-source. However, you need a Claude Cowork subscription (part of Claude Team at $25/user/month or Claude Enterprise at custom pricing) to use it. If you’re already paying for Claude Team, there’s no additional cost.
No. The plugin analyzes documents you provide but has no access to verified legal databases. For research requiring case citations, statutory interpretation, or regulatory guidance, you still need Westlaw or Lexis+. Claude will confidently hallucinate legal authorities.
Harvey is a production-grade platform with law firm-specific features: matter-level security, on-premise deployment, legal-specific fine-tuning. Anthropic’s plugin is a starting point that demonstrates Claude’s capabilities. Different products for different needs.
Basic setup requires comfort with Claude Cowork and MCP integrations. Customizing playbooks to match your company’s negotiation positions requires modifying configuration files. Most legal departments will want IT or legal ops support for implementation.
Claude Cowork has enterprise-grade security, but you should confirm data handling practices meet your privilege requirements. Anthropic’s enterprise agreements address confidentiality, but this needs review by your ethics counsel before processing privileged materials.
Probably not immediately. The plugin handles routine work (NDA triage, basic contract review) that overlaps with some CLM functionality. But it lacks the portfolio analytics, compliance features, and verified data that enterprise legal tools provide. Your existing stack likely does things this plugin can’t.
Last updated: February 2026. Based on Anthropic’s public announcements and open-source plugin documentation.