Latest in AI
Claude for Legal Artifacts: Contract Review, Cite Checks and Harvey Intake
Anthropic’s legal gallery puts reusable tools inside Claude, while Harvey shows why routing work back to the legal team matters as much as generating an answer.
Anthropic added ready-to-use legal artifacts to Claude on October 28, covering tasks including citation checking, contract review and regulatory tracking. Legal teams can adapt the interactive tools with their own documents and connectors, then share them with colleagues instead of rebuilding a workflow in every conversation.
This extends May’s Claude for Legal launch with workflows, not a new model.
What Claude for Legal artifacts do
The launch gallery includes a citation checker, a jurisdiction comparison matrix, a matter tracker, a contract playbook deviation board and a regulatory horizon scan. Artifacts open with examples and link outputs to their sources. Anthropic describes Claude doing a first pass, with a lawyer applying judgment afterward.
The distinction is useful. A reusable interface can make the same review procedure easier to repeat, but it does not establish that the procedure is reliable. Source links make checking possible; someone still has to determine whether the cited material supports the answer and applies to the situation.
For a team adopting these tools, the practical question is which parts of a review can be made consistent. A defined input, a visible output and an explicit review step are more manageable than a free-form conversation whose assumptions change each time.
How Harvey Intake connects Claude to a legal team
Harvey's launch announcement explains one concrete integration. The Harvey Intake artifact connects Claude to Harvey through its MCP server. Routine questions use the legal team's guidance, playbooks and prior work; requests needing legal judgment are routed to a lawyer in Harvey, with the response returned to the requester in Claude.
Requests are logged in Intake, where the legal team can assign and track them. Business users can check status and follow up without leaving Claude. This announcement adds the Claude artifact connection; it should not be read as the first launch of Harvey's broader Intake product.
Consider a sales colleague who needs help with a changed contract. The useful outcome is not simply a plausible explanation of the clause. The request needs to reach the appropriate reviewer, preserve the context and return an answer the business is authorized to use.
The handoff is part of the product. An assistant that can answer a question is not necessarily the system that should approve what happens next. Harvey's design makes that boundary more visible by keeping legal review in the legal team's environment.
What is available, and what remains unclear
Partners include Harvey, Goodwin, Foley & Lardner and PwC. Legora and Orrick additions are still forthcoming. The announcement does not provide a single standalone price covering the gallery and every connected partner service.
Teams should therefore check the requirements of the specific artifact they want to use, rather than assume that access to Claude includes every underlying system. A template is only one part of a working deployment; permissions, source access and responsibility for review remain separate questions.
This release also does not establish an accuracy benchmark for legal work. Its immediate value is a more structured way to package and share a process, not evidence that professional oversight can be removed.
What changes for enterprise AI workflows
A shared artifact gives a team something more concrete to inspect than a useful prompt buried in somebody's chat history. It can make the intended steps and outputs easier to discuss, revise and teach.
The harder test comes after that. Does the tool send uncertain cases to the right person? Can reviewers trace an output back to the input that produced it? Can the business tell the difference between an automated first pass and an approved answer?
Those are the questions that determine whether a legal AI workflow becomes part of normal operations or remains a convincing demonstration.
I’m Paras. I work at the intersection of AI, engineering and distribution.
I help technical teams turn product launches and complex workflows into clear stories that developers and founders can understand and share.