Crowdmines.ai Launches Research Helper MCP, the First MCP with Direct Answering Capabilities and Validated Results

TORONTO, Canada — November 4, 2026 — Crowdmines.ai, the agentic research analytics platform, today announced the launch of the Research Helper MCP. Research Helper is already available in the Crowdmines core platform; the Model Context Protocol (MCP) version now lets researchers connect it to their favorite LLM platform, such as Claude or ChatGPT. It is also the first MCP to turn a plain-language research question into a complete, multi-step analysis. Researchers ask a question; Research Helper identifies the right variables, plans and runs the sequence of cross-tabs and variable modifications needed, applies the correct significance tests, and returns downloadable cross-tabs, on weighted or unweighted data.

Survey analysis has traditionally meant hunting through hundreds of variables, building banners by hand, and waiting on a tab specialist. Research Helper removes those steps. A researcher can ask, for example, “How does purchase intent differ by age and region?” and Research Helper plans the analysis itself: it selects the variables, builds the banner, recodes where needed, runs the tests, and returns a finished, formatted cross-tab. The MCP version brings three advantages for researchers: they work in the LLM platform they already use; they get answers faster, because a question asked in conversation goes straight to a finished analysis; and they can shape the output to the deliverable they need, including dashboards and PowerPoint reports.

For research developers, Research Helper MCP is also available through a command-line interface (CLI). Developers can script and schedule analyses, rerun the same workflow on each new wave of a tracking study, and build Research Helper into existing data pipelines. Agencies and research technology teams can use it to deliver repeatable solutions to their clients, such as automated tab books, scheduled dashboards and PowerPoint reports, and self-serve portals. The same hard-coded formulas and validator agent sit behind every run.

What Research Helper MCP does

• Upload or reuse your data. Upload a survey dataset and questionnaire directly from the MCP, or work with data you have already uploaded to the Crowdmines platform.
• Plans the analysis from the question. Researchers ask in their own words, with or without variable names or table specs. Research Helper generates a multi-step plan: which cross-tabs to run, which variables to modify, and in what order.
• Cross-tabs and weighting. Run cross-tabs on weighted or unweighted data, including RIM weighting. Weight variables are built on the Crowdmines core platform, and the MCP uses them whenever they are available. Research Helper identifies the correct significance-test formula for each table, and every table can be downloaded for reporting.
• Charts and key findings. Results come with auto-generated charts and a summary of the findings that matter, ready to drop into a deck.
• Automatic variable modification. Research Helper searches the dataset’s variables and derives what the analysis needs on its own: combined variable levels, Top-2-Box scores, NPS, and other recodes.

How the agentic workflow works

Research Helper works through a research question like an analyst, planning and carrying out each step itself:

1. Interpret and find variables. It works out what to compare and selects the matching questions in the dataset.
2. Plan the steps. It decides which cross-tabs to run, in what order, and which variables to change first.
3. Modify the variables. It combines levels and calculates Top-2-Box, NPS, and other recodes.
4. Run the cross-tabs. It runs weighted or unweighted data through the correct significance test, using formulas hard-coded in the Crowdmines platform, not AI-generated code.
5. Validate, then deliver. A validator agent checks the results before anything goes out. Then Research Helper returns the tables, charts, and key findings, with the tables available to download.

Because every formula is hard-coded, no figure comes from AI-generated code or from the language model itself. Each count, percentage, and significance result is calculated by the Crowdmines platform and traces back to a table.

Security is built into the connection. Researchers sign in with their Crowdmines account, including multi-factor authentication, choose the organization they are working in, and approve permissions. Every read is authorized per request, and file contents go directly to Crowdmines rather than through the language model.

How the MCP version fits with the core platform

Research Helper already runs in the Crowdmines core platform. The new MCP version brings the same analysis engine into the tools researchers already use, and the two work side by side.

• MCP version: New, and built for speed without compromising calculation. It handles long-running tasks, so researchers can ask a whole list of questions at the same time. It can also use a weight variable built on the core platform whenever one is available.
• Core platform: Remains the home for the most thorough analysis. Its insight generation agent, available only on the core platform, takes additional time to deliver robust, calibrated responses that cover edge cases, and it is where researchers build weight variables.

“Research Helper works as a data science partner with market research experience. Researchers can still write every cross-tab spec and explore one table at a time, but now they can also ask the research question and get the answer directly,” said Shams Tabrez, Co-founder of Crowdmines.ai. “We built it to be a primary tool for data exploration and insight generation. It plans the cross-tabs from the research question, every formula behind it is hard-coded, and a validator agent checks the results before anything goes out.”

Built for how researchers work

Research Helper is designed for insights teams, agencies, and analysts who need rigorous, reproducible output. Rather than waiting for the user to specify each variable and table, Research Helper starts from the research question. Researchers can still write every cross-tab spec and explore tables one at a time; the Research Helper workflow adds the ability to answer the research question directly, like a data science partner experienced in market research.

What’s next

Crowdmines.ai continues to expand Research Helper. Open-end processing, including sentiment analysis and coding through topic modeling, is coming soon to the platform. Advanced analytics capabilities are also on the way. Research Helper will also add what-if answering, so researchers can ask how results would change under different scenarios.

Availability

The Research Helper MCP is available today. For details, visit the Crowdmines product page at https://crowdmines.ai/mcp-server, and see the how-to page at https://crowdmines.ai/resources/how-to/crowdmines-mcp to learn how to use it.

###