Adobe Customer Journey Analytics
Run cross-channel reports and build segments in Adobe Customer Journey Analytics.
Search global trademark registers, run knockout screens and pull full records.
Opens Agent Studio, where connecting is one click. The connector URL below works in any MCP client.
Paste it into any MCP client. Setup docs
The Clarivate IPOne CompuMark connector searches global trademark registers from Claude, ChatGPT or any MCP client. Nine tools run identical knockout screens, build fielded searches with phonetic matching, count result volumes before retrieval, and pull records and goods text. Every tool reads; none writes to any register.
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.
Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.
scopes_supported as an empty array and the authorization server advertises only openid, email and profile. Nothing at the OAuth grant narrows anything. On a wholly read-only server this costs less than it would elsewhere, but it also means a token cannot be scoped down.readOnlyHint or destructiveHint is reported here and parameter types come from Clarivate's documentation rather than from the wire.trademark-content caps at 100 IDs per call. Larger result sets must be batched by the client, and Clarivate does not document a cap on how many calls that may become.trademark-goods takes a single ID, so reviewing goods across a result set is one call per mark.WO, or EM for EU countries, produces a search that under-returns without erroring. Clarivate documents the rule; the tools do not enforce it.streamable-http, which is what our live probe exercised successfully. Follow the directory value.ipone.clarivate.com — including robots.txt, sitemap.xml and llms.txt — returns 401 with "Missing Bearer token". No crawl policy or machine index is readable for that host.No. All nine tools read. Their verbs are search, count, content, goods, fulltext and three that list valid field names, operators and office codes. No name carries file, create, update, delete, oppose or withdraw, and Anthropic's directory records the connector's permissions as read only. Nothing in this surface touches a register.
Clarivate does not say on this page, and that is the finding. CompuMark is sold by subscription rather than published list price, and the connector's documentation names no credit, quota, per-search fee or rate limit. No tool reports a balance or usage either, so an agent cannot check consumption before or after a search.
Nine, and this is one of the rare listings where both sources agree exactly. Anthropic's directory and Clarivate's own documentation page each name the same nine tools, with no name in one absent from the other. The directory prefixes each with ip-one___ where the documentation does not; the trailing segment matches in every case.
None that narrow anything. Its protected-resource descriptor publishes scopes_supported as an empty array, a deliberate statement rather than silence. The authorization server advertises only openid, email and profile, which are identity claims. Because every tool is read-only, no boundary is missing between a safe operation and a consequential one here.
Whichever ones you name. Every search tool requires a registrationOfficeCodes array in WIPO two-letter format, and trademark-search-register-office-codes lists the valid values. Clarivate advises adding WO with limitWOresultsToDesignated set true when searching one country, and adding EM as well for EU member states.
An exact-match screen. trademark-knockout-search looks for identical marks only and returns matching trademark IDs, which is the fast first-pass check for whether a name is plainly taken. It is narrower than trademark-search, which adds phonetic matching, plurals, cross-references and regional phonetics for Central European and Japanese marks.
Because search and retrieval are deliberately separate calls. The search and knockout tools return IDs only, then trademark-content resolves up to 100 of them per call into registration numbers, statuses, classes, dates and applicants. Goods and services text needs a further trademark-goods call, one ID at a time.
Yes. The endpoint returned an authentication challenge to an anonymous request on 2026-08-22, and Anthropic's directory records the server as requiring auth. Sign-in runs through Clarivate's own authorization server on ipone.clarivate.com, so access follows your existing IPOne CompuMark entitlement rather than a self-service signup.
WO/EM jurisdiction rule and the canonical-link instruction all come from this page. It is the documentation URL Anthropic's directory nominates for this connector. · retrieved 2026-08-22initialize request to https://ipone.clarivate.com/trademarks/mcp returned HTTP 401 with a www-authenticate Bearer challenge naming its RFC 9728 descriptor. · retrieved 2026-08-22scopes_supported present and empty, authorization server an Amazon Cognito pool (retrieved 2026-08-22) · retrieved 2026-08-22openid, email, profile; authorization code with PKCE S256. · retrieved 2026-08-22robots.txt (retrieved 2026-08-22). Permits ClaudeBot and GPTBot explicitly; no Content-Signal line. ipone.clarivate.com/robots.txt returns 401 and is unreadable. · retrieved 2026-08-22permissions: Read only, auth_posture: auth_required, transport streamable-http. · retrieved 2026-08-16Connect Clarivate IPOne CompuMark Trademarks once and your agents call these tools on their own: on a schedule, in a workflow, with nobody at the keyboard.