Lawve AI
by Lawve AI
Search a public catalog of practitioner-written legal Agent Skills from inside an AI agent. Two read-only tools, one marketplace:read scope, and no reach into case files, client documents or filing systems. Anthropic lists 2 tools; Lawve publishes no tool list.
Verified connector
Listed by Anthropic as a partner connector in its Connectors Directory.
Connection checked by Agentman on .
Anthropic states this reflects the level of review a connector received, not a security audit.
Connect Lawve AI via MCP
https://mcp.lawve.ai/mcpWorks in any MCP-compatible client. In Agentman Studio it is one click — no config file to edit.
Lawve AI Tools & Capabilities (2)
Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.
Limits
- The documentation URL Anthropic publishes for this connector is dead. The directory nominates
https://lawve.ai/en/learn/mcp-setup; that path returned 404 on 2026-08-23, as didllms.txt,llms-full.txtand the.mdsuffix on the same host. No/learn/route exists anywhere in Lawve's 12,212-URL sitemap. The directory'sprivacy_policyfield is stale in the same way, pointing athttps://lawve.ai/en/privacy, which also 404s — so two directory fields carry the same dead/en/prefix, suggesting the site dropped a locale prefix after the listing was written. The server itself is fully alive — a dead doc URL and a dead server are different findings. The working sources are Lawve's own posts and legal pages, reached through the sitemap and linked below. - Lawve publishes no tool list for this server. The two names come from Anthropic's directory. We could not diff them against a vendor enumeration because none exists. What Lawve does publish is a capability description — search the catalog, apply a skill, inspect the methodology — which matches the two names semantically but would not reveal a rename.
- We could not read tool schemas or safety annotations. The endpoint returned 401 to an anonymous request, so no parameter-level detail is published here.
- A prompt is advertised and could not be read. The directory lists
available_skills; the authenticated endpoint prevented anyprompts/listcall, so its real state is unknown. - Retrieved skill content is untrusted input. Manifests are community-authored instructions entering an agent's context. Lawve's security audit is a review, and Lawve's terms state review is not endorsement.
- Search quality is unverified. We never executed a search, so we cannot say how well the catalog's ranking surfaces the right skill, nor whether results are jurisdiction-filtered by default. Lawve names jurisdiction as a search dimension but publishes no ranking detail.
- No rate limits are published. We found no documented request ceiling for the MCP endpoint. Lawve's terms prohibit bypassing rate limits, which implies limits exist without naming them.
- Skills are not legal advice, by the vendor's own first line. Anything retrieved is a starting point for professional review. This connector does not change who is responsible for the work.
- No metering was found. Lawve states it is currently available without a subscription fee, and no tool spends money, credits or quota. There is correspondingly no accounting tool, which is unproblematic only for as long as nothing costs anything.
Frequently asked questions
No. Both tools operate on Lawve's public skills catalog, which is a library of published methodology documents, not a document management system. The connector holds no integration with a matter system, a DMS or a file store, and its single application scope is marketplace:read. Nothing in the surface reaches a client file.
Neither. Both listed tools are read-only, and the scope confirms it. The catalog does contain drafting skills, so a search result can return instructions that tell your agent how to draft. Your agent then does the drafting locally, using your own tools. Nothing is filed, sent or submitted through this connector.
Yes, in two places. Lawve's Terms of Use open with the sentence that Lawve does not provide legal advice, adding that output is not a substitute for professional judgment. Its Anthropic partnership post repeats that skills are starting points for professional review, not legal advice. Both statements are unusually prominent.
Nothing directly. We found no privilege waiver language and no privilege guarantee in Lawve's terms or privacy policy, so we record that as silence rather than as reassurance. The terms do require that content you publish does not unlawfully disclose confidential or privileged information, which addresses publishing rather than connector use.
Anthropic's directory lists two, and Lawve publishes no tool list of its own to check that against. The two names are lawvable_search_skills and lawvable_get_skill_manifest. Because no vendor enumeration exists, we cannot report a clean diff, only that there is no second list to diff against.
One application scope, marketplace:read. The server's RFC 9728 resource descriptor declared exactly that scope on 2026-08-23, and its authorization server advertises it alongside openid, offline and offline_access. The scope names read and no listed tool carries a write verb, so the grant and the tool surface agree.
Yes. Lawve's privacy policy describes the site as a public knowledge hub whose published resources can be viewed by anyone, and its skill pages load without signing in. The connector's value is bringing that catalog into a conversation, not unlocking private content. Sign-in is still required at the MCP endpoint.
Treat it as untrusted. The catalog is community-published, so a skill manifest is text written by a third party that arrives as agent instructions. Lawve runs a security review and shows a pass or fail badge per skill, but review is not endorsement under its own terms. Read a manifest before acting on it.
Sources
- Lawve AI, "Skills, Connectors, and Plugins: What Is the Difference?" (retrieved 2026-08-23) · retrieved 2026-08-23
- Lawve AI, "Lawve AI x Anthropic" partnership announcement (retrieved 2026-08-23) · retrieved 2026-08-23
- Lawve AI Terms of Use, July 2026 (retrieved 2026-08-23). Source of the "does not provide legal advice" opening line and the review-is-not-endorsement statement. · retrieved 2026-08-23
- Lawve AI Privacy Policy, July 2026 (retrieved 2026-08-23). Source of the public-knowledge-hub characterisation. · retrieved 2026-08-23
- Lawve AI connectors catalog, browse filters and counts (retrieved 2026-08-23) · retrieved 2026-08-23
- Lawve AI sitemap, used to establish that no
/learn/route exists (retrieved 2026-08-23) · retrieved 2026-08-23 - Lawve AI
robots.txt,Allow: /with noContent-Signalline (retrieved 2026-08-23) · retrieved 2026-08-23 - Lawve AI skill security auditor, describing the ten static check categories (retrieved 2026-08-23) · retrieved 2026-08-23
- Patrick Munro, "Lawve AI Becomes Anthropic Connector Partner" (retrieved 2026-08-23) · retrieved 2026-08-23
- Live OAuth posture check: anonymous initialize returning HTTP 401 with a
www-authenticateheader naming the resource metadata document; RFC 9728 metadata athttps://mcp.lawve.ai/.well-known/oauth-protected-resource/mcp; authorization server metadata athttps://auth.lawve.ai/.well-known/oauth-authorization-server(2026-08-23) · retrieved 2026-08-23 - Anthropic Connectors Directory entry — , read from our directory snapshot dated 2026-08-16 · retrieved 2026-08-16
- Lawve AI support — <mailto:contact@lawvable.com> · Privacy — . The directory's
privacy_policyfield points athttps://lawve.ai/en/privacy, which 404s on the same dead/en/prefix as the nominated documentation URL; the working page is linked here instead.
Use in Agentman
Connect once and your agents call these tools on their own — on a schedule, in a workflow, with nobody at the keyboard.
Open in Agentman StudioServer Info
- Category
- Productivity
- Developer
- Lawve AI
- Tools
- 2
- Domain
- mcp.lawve.ai
Using Claude Desktop or another MCP client? Setup docs — the connection URL above works anywhere.