Anthropic announced a Claude for Legal expansion Tuesday, adding plugins for specific areas of law โ commercial, privacy, corporate, employment, product, and AI governance โ plus MCP (Model Context Protocol) connectors to DocuSign, Box, and Thomson Reuters Westlaw. The new features are available to all paying Claude customers, not gated behind enterprise sales the way Harvey and Legora gate their vertical-AI products. The competitive backdrop: Harvey raised $200 million at an $11 billion valuation in March, and Legora raised a $600 million Series D last month with a Jude Law-fronted ad campaign. Anthropic is moving directly into a vertical the AI-startup ecosystem already calls "heating up."
Two architectural moves matter here. First, the plugins-by-practice-area structure means Anthropic is shipping pre-built workflows for clerical tasks โ document search and review, case law research, deposition prep, document drafting โ tuned to specific legal domains rather than relying on a generic LLM endpoint. Second, and more important, the MCP connectors let Claude integrate directly with the software law firms already use: DocuSign for document management, Box for file storage, Thomson Reuters Westlaw for legal research. MCP is Anthropic's open protocol for agent-to-system integration, and it's the primitive driving the broader agent stack this year (Signadot's K8s validation skill, Manifold's MCP server scoring, AWS Kiro's neurosymbolic pipeline all sit on or near it). Anthropic shipping legal-domain MCP connectors is a distribution play: the existing law-firm software keeps its surface area, Claude rides on top via MCP rather than replacing the stack a partner already paid for.
The vertical AI legal market has been Harvey, Legora, and a handful of smaller firms (Casetext, Hebbia, EvenUp). Their pitch: domain expertise, fine-tuned models, enterprise sales. Anthropic's pitch is structurally different โ foundation-model breadth, MCP-connector distribution, included with the Claude subscription law firms already have for other knowledge work. This is the same shape as Microsoft 365 Copilot eating into vertical productivity SaaS: when the platform vendor ships the integration layer, the vertical specialists have to compete on something other than integration. For Harvey and Legora, the durable answer presumably becomes deeper workflow automation, specialized eval harnesses, and the legal training data they've acquired that Anthropic hasn't. For Anthropic, the calculation is that "good-enough legal with great Claude integration" beats "great legal with mediocre integration" for most law firms most of the time. The OpenAI wrongful-death suit covered separately today (#827) sits in the background โ the "unauthorized practice of medicine" legal theory has obvious cousins for "unauthorized practice of law" if AI legal tools start advising clients directly. Anthropic's framing here as "automating clerical functions" rather than "providing legal advice" is deliberate liability positioning.
Available now to all paying Claude customers, plugins shipping across the six practice areas listed, MCP connectors to DocuSign/Box/Westlaw immediately. For law firms: this is a make-or-buy moment for AI legal infrastructure โ Harvey/Legora pricing versus existing Claude subscription with these plugins included. For Harvey and Legora: the competitive question becomes whether their specialized tuning and training data delivers value the foundation-model + MCP stack can't replicate. For everyone else watching frontier-lab strategy: this is Anthropic's second major vertical move (after Claude for Education), and it suggests the pattern of "foundation model with domain plugins + MCP connectors" is the distribution shape Anthropic is committing to. Healthcare is the obvious next vertical to watch โ and the OpenAI wrongful-death suit puts a brake on how aggressively any frontier lab can push into health advice without explicit "clerical only" positioning.