Connect your spatial data
Upload the files already sitting on your shared drive, or connect to the systems your organisation publishes from. mapbench sorts out the differences between them and puts everything on the same map.
AI-driven spatial analytics
The workbench for building maps with AI agents.
Drive a powerful spatial agent harness. With a full toolbox available for geospatial analysis. Bring your spatial data, import spatial services from the web and build mapping solutions that can be shared across your organisation.
How it works
Ask in plain language
Connect Claude and ask for what you need. It works with your real data rather than a summary of it, does the analysis in your workspace, and shows you how it got to the answer.
Share the analysis
Every answer is a real map, not a screenshot. Style it, send someone a link, or drop it into a page you do not own — and take the link back whenever you want.
Working with your AI agent
A GIS analyst in your laptop
Connect your AI agent to your mapbench workspace and let it direct your spatial analytics projects.
Point it at a source and it imports. Ask it a question and it writes the query, builds the model, and turns the result into a layer. Ask for a map and it builds one: layers stacked, styled, labelled, ready to share.
You work from the chat client you already have open. Nothing gets exported, nothing gets copied, and the agent's edits land in the same history as everyone else's.
Geospatial projects
- Which of our sites sit inside the new flood area?
- How many homes are within 500 metres of this route?
- Where are we furthest from a depot?
- What has changed around here since last year?
- Put everything we manage on one map.
- Claude
- Cursor
- GitHub Copilot
- Windsurf
- Zed
- Cline
- Gemini
Build models
Chain models together like building blocks.
A model is a saved SQL step over your datasets and the shared reference layers — a spatial join, a filter, a rollup. Its output is data in its own right, so the next model builds straight on top. Snap them into a pipeline as deep as the question needs, and reuse every piece.
Start from the data you have
Point a model at your own datasets and the shared reference layers, describe the relationship in SQL, and mapbench runs it in your workspace and turns the result into a layer.
Every model is a new block
A model's output is a dataset in its own right, so the next model builds on it. Compose small, legible steps instead of one query nobody can unpick — and change one without rewriting the rest.
Or let your agent build the chain
Ask your connected AI agent for the end result and it wires the models together, checks each step against your real data, and hands you the finished map — the same pipeline, built for you.
Built with security at the core.
- Default Privacy Nothing is public unless you choose to share it.
- Strict Workspace Isolation Keep teams, projects, and permissions cleanly separated.
- Full Audit Logging Every edit—human or AI—is timestamped with clear attribution so you always know who changed what.
- Instant Link Revocation Take shared maps offline in a single click.