See QMerse agents at work.

A scheduled QMerse agent captures public vehicle-market listings, structures vehicle and dealer data, maps market activity by ZIP code, and stores a daily snapshot for ongoing analysis. It runs on the server on a cron schedule — nobody opens the canvas to trigger it.

Track new inventory, regional pricing, price reductions and days-on-market — down to individual VINs.

How it is built. A Python script node handles collection and flattening, then writes straight into PostgreSQL. The write is configured to skip on conflict, so a re-run never duplicates rows and the table stays clean whether the flow fires once or fires again after a failure. The schedule is a standard cron expression, so the flow runs unattended and reports its own run history.

Why the schedule is the point. A single scrape is a snapshot; the value is in the accumulation. Because every run appends another dated snapshot, the dataset becomes a time series — and questions that are impossible against one download become straightforward: which ZIPs are discounting fastest, how long a given trim sits before it moves, where a specific VIN has been repriced.

The map shows listings by dealer ZIP, with bubble size as listing count and colour as average price. It is generated from the same workflow that collects the data, so the visualisation stays in step with the table behind it.

Could your Texas property be over-appraised compared with similar homes in your neighbourhood? This agent connects directly to property databases, analyses comparable homes, and helps a homeowner build an evidence-based tax protest.

End to end, it:

• Compares your appraised value and price per square foot against similar nearby homes
• Runs the mathematical and statistical analysis needed to detect over-appraisal
• Identifies the most relevant comparable properties
• Plots the subject property and its comparables on an interactive map
• Estimates a supported value, the potential assessment reduction and the possible tax saving
• Generates an evidence report with charts, maps and protest arguments
• Prepares the official Texas protest form for you to review and submit

The last step is the one that matters. Most AI stops at analysis and leaves a person to do the paperwork. This carries it through to a completed, ready-to-submit form and an evidence pack — not a summary of what you could do next. You review it and file it yourself; that is the difference between something interesting and something a homeowner actually uses.

Behind the front end, agents designed and orchestrated in QMerse manage the whole chain: database queries, calculations, analytics, mapping and automated report generation. It is a good example of a data-intensive process that nobody enjoys, turned into an application that produces a clear, actionable result.

This demonstration is for informational purposes and does not constitute tax or legal advice. Any generated analysis or form should be reviewed before submission, and a successful protest or tax saving is not guaranteed.

An agent connected to a Houston housing database, deployed as an application people can actually use. You ask in plain language — “Show me houses in the Houston area”, “Highlight homes priced below the median” — and the agent analyses the data, returns the results, and updates the interactive property map in real time.

The demo shows both halves at once. On the left is the QMerse visual workspace, where the agents, tools, data inputs and workflows are designed and orchestrated. On the right is the user-facing application: the conversational interface and the property map your end user actually touches. You can watch a request travel from the canvas that was built to the screen a customer is looking at.

What the agent coordinates behind that interface: understanding the request, querying the database, filtering the results, and controlling the map. Note the last one — the agent is not just returning rows for someone else to render. It is operating the application.

The pattern generalises well beyond property. Most organisations have a database that only two or three people know how to query, so the data is available in principle and unreachable in practice. This is the shape of the fix: an agent that can talk to a database, drive an application, and control a visual interface on the user’s behalf.

A critical-minerals intelligence application powered by agents designed and orchestrated on the QMerse canvas. It connects directly to multiple geological and mineral databases, and users talk to that data in plain language while the agent runs the analysis and takes actions inside the application.

What the agent does:

• Searches and analyses lithium brine wells and mineral deposits
• Identifies and ranks high-grade lithium opportunities
• Compares concentration, depth, location and other technical attributes
• Explores regions such as the United States and the South American Lithium Triangle
• Combines information across several geological datasets at once
• Filters, navigates and updates the interactive map automatically
• Supports early-stage technical and commercial feasibility assessment
• Saves selected opportunities and generates structured reports for investors

The front end is a real application — an interactive geospatial map, a database explorer and a conversational interface. Behind it, QMerse agents coordinate the database queries, the analytical tools, the map actions, the visualisation and the reporting. The agent does not just answer; it operates the application on your behalf.

This is what turns a complex geoscience dataset into a decision-support system that explorers, mining companies, investors and government bodies can all use — without each of them needing to know how to query it.

This application supports preliminary screening and decision-making. Its outputs do not constitute geological, financial or investment advice, and any opportunity should undergo independent technical and commercial due diligence.