GXLab developer tools
Your assistant. Our documentation.
Connect the VASAT MCP and Claude or ChatGPT writes code against the real GXLab platform. The right model. The right fields. The right query. First time.
https://accounts.vasat.io/llm/mcp
Stop pasting documentation into chat windows.
The docs come to your assistant. Ask a question in plain language and get an answer drawn from the live platform, with the source section named so you can check it in seconds.
Everything GXLab runs on.
Three platforms, one connector. Name the one you want and the search goes straight there.
The GXLab SDK
Farms, samples, points, labs, batches, boundaries, harvest jobs and raster layers. The product layer, and where most work starts.
The framework
Sessions, OAuth, query building, filter syntax, paging, GraphQL and path-based access control. The deepest coverage of the three.
The Spatial Engine
The Geospatial Management Server. Filesets, folders, vector data and point clouds, with full object models.
Built for people who ship.
Six reasons it earns a place in your toolchain.
The right model, first time
The correct model, its filterable fields and the query that reads it. Before you write a line.
Code you can run
Real samples in TypeScript, Shell, JSON, Scala and HTML, drawn from the working platform.
Read your own data
Describe what you want in plain language. Your assistant writes the query. You run it under your own credentials.
Answers you can verify
Every claim traces back to a named document section. Ask for the source and check it yourself.
Nothing to install
A hosted server and a URL. No packages, no local setup, no maintenance.
Your data stays yours
The server holds no credentials and no session. It serves documentation. Your records never leave GXLab.
Connect it in a minute.
Pick your assistant. Paste one URL. Start building.
Open Settings, then Customize, then Connectors.
Click the plus button, then Add custom connector.
Name it VASAT and paste the server URL.
https://accounts.vasat.io/llm/mcpClick Add. That is the setup done.
Switch it on in any chat with the plus button, then Connectors, then VASAT.
Mobile: connectors added on the web are ready to use in the app. Desktop: same Settings flow as the web.
Run one command.
claude mcp add --transport http --scope user vasat https://accounts.vasat.io/llm/mcpConfirm it registered.
claude mcp list
Want it committed to the repo so the whole team gets it? Use --scope project. That writes .mcp.json:
{
"mcpServers": {
"vasat": {
"type": "http",
"url": "https://accounts.vasat.io/llm/mcp"
}
}
}Open Settings, then Connectors, then Advanced, and turn on Developer mode.
Return to Connectors and choose Create.
Name it VASAT and paste the server URL.
https://accounts.vasat.io/llm/mcpLeave authentication at the default and save.
Select it per conversation with the plus button, then Developer mode.
See it work.
Open a chat with the connector on and ask this.
Using the VASAT connector, search the gxlab docs for how to fetch paddock boundaries. Name the document section you used.
You get the method, the model it belongs to, and the document section behind it. Boundary in GX Models, or recipe 4.4 of the Integration Guide. Named, so you can check it.
What people are building.
Real work, already shipped on the platform. The connector gives your assistant the models, methods and field names, so the build starts at the interesting part.
Paddock boundary viewer
A live map of a farm's boundaries in Leaflet or similar. Draws on Boundary in GX Models.
Consolidated soil report
Every sample for a farm, grouped by depth and analyte, rendered to PDF. Draws on Dropzone Models.
Soil test chat assistant
A chat window over a customer's own results, with an LLM summarising them in plain language.
Carbon and emissions calculators
SOC results and paddock areas fed into a methodology model. VM0042, VT0014 or your own.
Raster layer viewer
Harvest output surfaces for clay, carbon and yield, rendered over a boundary.
Job status dashboard
Every processing job for an account, tracked in one view. Draws on Harvest Jobs Overview.
Entitlement and usage check
What an account is subscribed to and how much of its allowance is used. Draws on Shepherd Models.
Satellite layer browser
Every raster layer available over a boundary, listed and rendered. Draws on Tiff Selectors.
Your data, your machine.
The connector serves documentation. Your records stay in GXLab.
Describe what you want in plain language. "Every soil sample for farm 1730 with carbon results, grouped by depth." Your assistant searches the models, comes back with the query, and you run it under your own credentials. Nothing sensitive crosses the wire.
Say what you want in domain terms.
Name the platform. Your assistant returns the model, the filterable fields and a working query.
Run it against a farm you know and confirm the result.
Widen it, schedule it, or build it into a product.
One URL. One minute.
Paste it into your assistant and start building against the real platform today.
Connect the MCP Talk to GXLab