.webp)
Thank you to everyone who joined Patent answers, inside the AI you already use. Across the two sessions on 4 November, you sent us more questions than we had time to answer live. We have grouped them by theme below, merged similar questions and included the answers given during the sessions.
The morning session, Foundations, covered when to use the MCP. John Paul Keeler and Yashasvi Ranawat were joined by Lennard Mooij of Brann and Olli Pekonen of Vensum Power, who showed how they use the IPRally MCP in their own work. The evening session, Advanced MCP, focused on repeatable workflows. John Paul walked through Tony Trippe's workflow at Owens Corning, Luis Limon of IPRally showed a skill that builds a full landscape report from one collection, and John Paul showed his own analysis of Apple's filings around the 2007 iPhone keynote.
Getting started
Q: What exactly is the MCP, and what does IPRally provide?
A: The IPRally MCP connects the AI assistant you already use, such as Microsoft Copilot, ChatGPT, Claude or an in-house model, to the IPRally search engine through your own IPRally account. Your assistant can run AI similarity searches, Boolean searches and full document reads, and it can reuse the collections, search cases, projects and agent runs you have saved in IPRally. The MCP also ships with built-in instructions, so the assistant knows which tool to use for each question without prompt engineering. IPRally provides the search and the data; your assistant does the reasoning on top.
Q: How does a self-built interface, like the one Lennard showed, relate to what IPRally offers? I'm not someone who builds software.
A: You do not need to build anything. The MCP works directly inside Copilot, ChatGPT or Claude, which is how John Paul ran the examples in the session. Lennard's invalidity tool is his own software, built on top of the MCP so he can choose his model and the format of his claim charts. His advice for anyone curious was to open the Code tab in Claude, describe what you want to build and let it guide you.
Q: How do I get access to the MCP?
A: The MCP is included with IPRally Professional and Lite licences, subject to approval from your IT team. Professional users get every tool. Lite users get up to 100 results per search and Boolean filters on smart search, but not standalone Boolean search or tag filters. Existing customers can contact their account manager or use the support chat in IPRally. New to IPRally? Request a demo.
Q: Where can I find instructions for connecting my AI assistant?
A: Our Help Centre has setup guides for each assistant. ChatGPT users can add IPRally from the ChatGPT plugin directory. For Microsoft Copilot in an enterprise setting, your IT team makes the connection, and we are happy to help them through it.
Q: Does the email for my Copilot account need to match my IPRally account?
A: No. The two accounts are separate. When you connect, you sign in once in your browser with your own IPRally account, and that login decides what the MCP can access. If your company uses single sign-on, contact your account manager so we can make sure your domain is set up.
Q: Some of us use Copilot rather than Claude. Does it work the same?
A: Yes. The MCP gives every assistant the same tools: find patents, run a Boolean search, read a document. The search results come from IPRally whichever assistant you use. What differs is the analysis on top, which depends on the model behind your assistant. As John Paul put it, the MCP works the same way everywhere, but the output depends on the engine you connect it to.
Data, security and cost
Q: How can I make sure the AI doesn't access sensitive novelty search projects? Can I choose which projects it can see?
A: The AI assistant reads a saved search, collection or project only when you point it there, for example with Export > "Copy as prompt". It does not go looking through your work on its own, though there is no setting today to limit it to specific projects. It sees your saved work under the same sharing rules as the IPRally app, so it cannot reach anything your account cannot. Your assistant sends IPRally only the search query, never the chat, and we do not store MCP searches. If a future version lets your assistant save work back into IPRally, it will ask your permission first. IPRally data is encrypted at rest, and ISO 27001 covers the MCP through our infrastructure.
Q: Will there be additional token requirements for IPRally search?
A: Not on the IPRally side. MCP access comes with your licence at no extra cost. To keep usage fair, each user has a rate limit on requests per second, and your assistant will tell you if you reach it. Your AI assistant uses tokens as normal under your own AI subscription. To keep that low, IPRally returns lean results by default (titles, dates and scores) and adds detail only when a workflow needs it.
Q: If I ask the same question twice, will I get the same answer?
A: The search results will be the same, because they come from the IPRally search engine. The reasoning your assistant writes on top depends on the model you use, so the wording and conclusions can vary. As Olli put it in the session, if you depend on the result, double-check it.
Features and roadmap
Q: For R&D it would be very useful to see patent images through the MCP. Will this be possible?
A: Not today. Images make each response much larger and take up the assistant's working space, and a full patent document is already large. Every result includes a link to the document in IPRally, where you can view the figures.
Q: Will Analyst AI, a chat-with-a-set function, be part of the MCP?
A: Analyst AI is not part of the MCP. You can get much of the same result through the MCP already: in IPRally, choose Export > "Copy as prompt" on a collection, search case, project or agent run, paste the prompt into your connected assistant and ask your questions about that set. John Paul showed this in the session with an agent run he shared with an engineering team.
Q: Can colleagues without their own IPRally licence chat with the results of my search or collection?
A: Not yet. Each person connects with their own IPRally login, and colleagues with a licence see shared items under the usual sharing rules. If you want to bring wider R&D teams in, talk to your account manager.
Q: Can the agent do an inventive step attack using the EPO problem-solution approach or the USPTO approach?
A: Both guests have tried this. Lennard said current frontier models handle these attacks, and defences, well under either approach, and that they have improved a great deal in the past year. Olli added that for a single publication you can also ask the question directly in IPRally, without the MCP.
Q: Will you share the mobility example for inspiration?
A: Yes, here is what John Paul asked Copilot in a single prompt: to explore the robotaxi and mobility space, find the closest patents to the idea, compare named competitors filing since 2020, review the claims of the two closest documents, and check them against a collection in an existing IPRally project.
Q: What should we build next?
A: We ran a poll in each session. In the morning, an FTO claim overlap report was the most requested, with 47 of 94 votes, followed by monitoring alerts through the agent. In the evening, saving agent results back into IPRally projects and collections came top, and monitoring alerts were requested again. Thank you for voting. The results go straight to the product team.
The role of the specialist
Q: How will the role of the patent information specialist look in the future?
A: In the session, Yashasvi described two kinds of work. When the search is the work, as in a novelty search, an FTO search or a landscape, the specialist runs it in IPRally, reviews it, tags it and defends it. When the search feeds a bigger task, such as comparing a roadmap to competitors' filings, the MCP lets others build on that work. John Paul added that this makes specialists more valuable, because the collections and tags they curate become the trusted base the rest of the organisation draws on.
As IPRally co-founder Sakari Arvela wrote in When systems start talking, human supervision and goal-setting become "more important, not less": "The machine handles the relentless, repeatable substance work; people handle judgment and direction."
From the Advanced MCP session
Q: In Tony's case study, who linked the patent features to the regulation requirements: Claude, IPRally or Tony?
A: Claude made the link, and Tony checked it. IPRally searched and compared both portfolios through the MCP and returned the patent results. Claude then combined those results with the regulatory requirements and the other public sources Tony gave it. Tony reviewed the output and made the final calls, as he did when the model suggested widening the set of competitors.
Q: Did Apple file utility models in Germany in 2007, and did John Paul's case study cover German utility models?
A: John Paul's case study counted US applications by filing date and followed the European family through priority chains. In the session, he pointed to German utility models as the route that helped Apple in Germany. A German utility model has a six-month grace period for the applicant's own disclosure, while the European patent route has none, and the swipe-to-unlock filing came just under six months after the January 2007 keynote.
Q: Will the recordings of both sessions be shared?
A: Yes, both are available on demand. Watch Foundations for when to use the MCP and the four core moves, and Advanced MCP for turning patent questions into repeatable workflows.
What's next
The MCP gives you IPRally search inside the AI you already use, with your saved work one prompt away. Our next webinar, Patent search agents in practice, on Wednesday 18 November, looks at the Novelty and Invalidity agents working through real cases and gives a first look at the FTO agent. Register here.
Want to see the MCP on your own questions? Request a demo, or contact your account manager if you are already a customer.
Further reading
- Your IPRally workspace, wherever you work: what the latest MCP release adds, from collections and tags to agent runs, and how to hand saved work to your AI assistant.
- When systems start talking: IPRally's journey toward a virtual IP department: Sakari Arvela on where the MCP and API fit, and why human judgment matters more as agents take on more of the work.
.webp)
