Cypris Q

By Cypris

Get New App Alerts

Open in ChatGPT

Open in Claude

Overview

Cypris Q brings specialized R&D and innovation intelligence to your team in Claude through a secure connection to your Cypris account. Unlock access to Cypris's global intelligence layer across patents, scientific literature, chemical compounds, companies, and technical markets. Your team can use Cypris Q inside Claude to investigate prior art, freedom to operate, competitive landscapes, technology trends, whitespace opportunities, and other critical questions across the R&D and IP stage-gate process. Ask a research question and Cypris goes to work, returning a cited report in the conversation a few minutes later. Unlike general-purpose AI tools, Cypris is purpose-built for high-stakes technical and strategic research. Responses are grounded in authoritative source material, with citations that allow users to review the evidence behind each finding. Cypris Q is designed for R&D, IP, innovation, strategy, and technical teams that want to make Claude more valuable by extending it with specialized datasets, research workflows, and domain-specific intelligence. Getting started - Add Cypris Q and sign in with your Cypris account when Claude prompts you. - Ask your research question. A run takes three to five minutes, and the report arrives in the conversation. - No Cypris account yet? Request access at https://www.cypris.ai/book-a-demo - Learn more about the Cypris integration at https://cypris.ai/products/integratedintelligence - Need help? Email accounts@cypris.ai Cypris Q accesses only data scoped to your Cypris account, and your data is never used to train AI models. SOC 2 Type II and US-based.

Tools

cypris_list_agents

ChatGPT

Use this function to list the Cypris agents this user can run: their own saved agents, any shared with their organization, and the Cypris built-ins. Each entry carries an id, a title, a description of what that agent does, and a scope of cypris for a Cypris built-in, organization for one shared with their organization, or personal for their own. The built-ins also carry a category -- innovation, competitors, patents, market or scientific_academic -- so narrow by category rather than reading every title. Call it when the user asks which agents or saved prompts they have, or names one and you need to know whether it exists. It takes no arguments and always returns that user's own visible set, so never ask the user for an id or an account to list. If truncated is true the list was cut short, so "not in the list" does not mean the agent does not exist.

cypris_list_runs

ChatGPT

Use this function to find a Cypris research run you no longer have the run_id for, so you can fetch its report or carry on answering it instead of starting the research over. Each entry has a run_id, the title the run was given from its opening question, started_at, and can_reply, which is false while that run is still working -- cypris_research_reply refuses a run until it is true. Call it when the user refers to research they asked for earlier and you have no id in hand. Pass search to match the title, or leave it empty for the most recent runs. It only ever returns this user's own runs.

cypris_research

ChatGPT

Use this function when the user asks a research question Cypris R&D intelligence can answer — patents, scientific papers, companies, funding, or the technology landscape of a topic. Pass the research question as prompt; Cypris runs a full agentic research process producing a cited markdown report. Starting it is ALL this function does: it returns a run_id at once and NEVER carries the report. A run takes three to five minutes. So when status is "running", do not wait or poll in this turn — end your turn, and pass that run_id to cypris_research_result when the user next asks. When status is "failed" no run started: report the error. Do not call cypris_research again to check on a run. You may run a few research calls in parallel for DIFFERENT questions, but never re-submit the SAME question while its run is active. To run one of the user's saved agents, pass its agent_id from cypris_list_agents; omit it for ordinary research.

cypris_research_reply

ChatGPT

Use this function to answer a Cypris research run that came back with questions instead of a report. Many Cypris agents ask what you are trying to do before they research anything, so a run started with agent_id often returns questions on its first result. Pass the same run_id and put the answers in prompt, then fetch with cypris_research_result as before. Answer from what the user has already told you where you can, and ask them only for what you genuinely do not know. You can reply more than once. Wait for the run to stop being "running" before replying again; a reply while it is still working is refused.

cypris_research_result

ChatGPT

Fetch a Cypris research run started by cypris_research: pass its run_id. When status is "complete", results carries the cited markdown report and its sources: each inline [N] marker refers to the returned source whose index is N. A "complete" run does not always carry a report: an agent that needs to know more comes back with its questions as the report content instead. When it does, answer them with cypris_research_reply and the same run_id. If status is "failed" the research ended without a report: tell the user it could not be completed, report the error, and do not fetch that run_id again — unless the error is "cypris_account_inactive": that report still exists, so fetch the same run_id once access is restored. Runs stay retrievable: a run_id can be fetched again later. Do not call cypris_research to check a run; that starts another one. If status is still "running", tell the user it is still going and end your turn — never call this function twice in one turn.