> For the complete documentation index, see [llms.txt](https://docs.geopard.tech/geopard-tutorials/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.geopard.tech/geopard-tutorials/product-tour-web-app/geopard-mcp/geopard-mcp-capabilities.md).

# GeoPard MCP Capabilities

## GeoPard MCP Capabilities

GeoPard MCP gives compatible AI assistants permissioned access to your GeoPard account. It works with your real farm data. It can analyze field variability, create operational outputs, and send them to machinery.

Use it to create zones maps, prescription maps, sampling plans, and machine-ready files. This reduces manual preparation time while keeping review and approval in your hands.

For setup, see [Connect GeoPard MCP](/geopard-tutorials/product-tour-web-app/geopard-mcp/connect-geopard-mcp.md) and [Verify GeoPard MCP Connection](/geopard-tutorials/product-tour-web-app/geopard-mcp/verify-geopard-mcp-connection.md).

{% hint style="info" %}
Every data tool runs under your OAuth session. The assistant accesses only farms, fields, and datasets available to your GeoPard user.
{% endhint %}

### Capability map

| Group                | What it does                                                              |
| -------------------- | ------------------------------------------------------------------------- |
| Discovery            | Find farms, fields, and field data layers, including topography           |
| Analytics            | Calculate vegetation statistics, terrain context, and agronomic equations |
| Zones and VRA        | Create zones maps and assign product rates                                |
| Yield data quality   | Clean, calibrate, and generate yield datasets                             |
| Sampling             | Create grid and zone-based sampling plans with routing                    |
| Export and machinery | Export files, push to John Deere, and create WorkPlans                    |
| Navigation           | Open precise GeoPard views for visual review                              |

### Discovery and context

The assistant first resolves your farm and field context. This grounds every answer in your data.

| Tool                       | Purpose                                                                                     |
| -------------------------- | ------------------------------------------------------------------------------------------- |
| `searchFarms`              | List and filter farms                                                                       |
| `searchFields`             | Find fields by farm, name, or filters                                                       |
| `searchSatelliteImages`    | Find field imagery within a date range                                                      |
| `searchSoilDatasets`       | Find soil lab and sensor datasets                                                           |
| `searchTopographyDatasets` | Find topography maps: DEM, slope, aspect, hillshade, TPI, TRI, and roughness                |
| `searchYieldDatasets`      | Find harvest datasets by field and season                                                   |
| `searchAsAppliedDatasets`  | Find machinery application records                                                          |
| `searchZonesMaps`          | Find zones maps, geometries, and assigned rates                                             |
| `searchEquationMaps`       | Find previously generated equation maps                                                     |
| `searchSamplingPlans`      | Find plans by field, type, status, or zones map                                             |
| `searchTrials`             | Find rate and variety trials                                                                |
| `searchProvider`           | Search organizations, farms, and fields and other data layers in John Deere, CNH, and AGCO. |

**Projection control.** Searches support `ids`, `summary`, `statistics`, and `full` projections. Use compact results for listings. Request full details only when geometries or attribute statistics are needed. This reduces round trips on large accounts.

### Analytics

#### Vegetation index statistics on any geometry

`searchSatelliteImageryByGeometry` calculates index statistics for a Point, MultiPoint, Polygon, or MultiPolygon over a date range. Use it to inspect a suspect area, compare zones, or assess trial strips.

* **Providers:** Sentinel-2, Landsat 4, 5, 7, 8, 9, and Planet.
* **Indices:** NDVI, EVI, EVI2, RVI, LAI, OSAVI, SAVI, GNDVI, IPVI, GCI, WDRVI, RCI, SBI, GFPI, NDMI, MSI, CCCI, MCARI, TCARI, MCARI\_OSAVI, TCARI\_OSAVI, MCARI1, NDWI, NIR, and NDYI.

#### Topography and terrain

`searchTopographyDatasets` finds the terrain layers available for a field. Use it to explain water movement, erosion risk, cold spots, and yield patterns that imagery alone does not show.

* **Layers:** DEM, slope, aspect, hillshade, combined hillshade, TPI, TRI, and roughness.
* **Filters:** farm, field, specific topography map UUIDs, and status `SAVED`, `EXECUTED`, `ERROR`, or `EMPTY`.
* **Projections:** `ids`, `summary`, `statistics`, and `full`, as with other searches.

A farm or field filter is required, because topography maps belong to fields.

Terrain attributes feed straight into `generateEquationMap`. Combine slope with soil and imagery to place lime and drainage work, reduce seeding on eroded ridges, or set variable tillage depth. Topography maps can also be pushed to John Deere as Maps Layers for in-cab context.

#### Equation-based agronomy

`generateEquationMap` runs transparent formulas across multiple layers and saves the result as a map.

* **Input variables** can use imagery indices, soil attributes, yield data, as-applied data, topography, zones maps, and equation maps.
* **Dry run** defaults to `true`. It validates syntax with sample values before generation.
* **NumPy support** enables conditional logic with functions such as `np.where` and `np.isnan`.

Choose a manual grid or an AB-line grid. AB-line grids use machine working width, headlands, sections, application length, and turning radius. This aligns prescriptions with actual field operations.

Supported purposes are `GENERAL`, `SEEDING`, `FERTILIZING`, `SPRAYING`, and `IRRIGATION`.

#### Geometry and raster access

| Tool            | Purpose                                                     |
| --------------- | ----------------------------------------------------------- |
| `fetchGeoJson`  | Retrieve a GeoJSON FeatureCollection from a GeoPard WFS URL |
| `getGeoMapUrls` | Resolve rendered map URLs for visual review                 |

### Zones and variable rate

| Tool                  | Purpose                                                                              |
| --------------------- | ------------------------------------------------------------------------------------ |
| `saveZonesMap`        | Create zones from Polygon or MultiPolygon geometries, with optional rates and colors |
| `assignZonesMapRates` | Assign up to three products' rates to an existing map                                |

Each product has a name, rate unit, and optional price per unit. One map can include MAP, urea, and seed rates. Supported units include metric and imperial mass, volume, seed counts, currency per area, and tillage depth.

{% hint style="warning" %}
`assignZonesMapRates` updates an existing map. The first call returns a rate comparison preview. GeoPard writes changes only after explicit confirmation.
{% endhint %}

### Yield data quality

| Tool                            | Purpose                                                        |
| ------------------------------- | -------------------------------------------------------------- |
| `cleanCalibrateYieldDataset`    | Clean and calibrate a yield dataset                            |
| `generateSyntheticYieldDataset` | Create a yield layer when measured harvest data is unavailable |

**Cleaning options** include auto-cleaning, attribute min/max filters, sigma filters, and the USDA protocol. The USDA workflow supports crop-specific yield, moisture, velocity, timing, swath, and overlap settings.

**Calibration options** can target a field average or total. They can clamp values to a defined range. They can also calibrate by machine path with adjustable smoothing.

Synthetic yield supports corn, corn silage, soybeans, wheat, cotton, rice, barley, oats, sorghum, canola, sunflower, rapeseed, and rye. Set a target average or total, harvest year, and date.

{% hint style="info" %}
Yield cleaning and calibration require a pre-flight check. Review dataset statistics and confirm ambiguous target attributes before corrections run.
{% endhint %}

### Soil sampling plans

`generateSamplingPlan` creates or updates `GRID` and `ZONE` plans for a field.

* **Placement:** GeoPard Core Line, Optimized Coverage, NZ, or W patterns.
* **Routing:** Zone-by-zone routing or smart ordering across zones.
* **Sampling:** Set points per zone, depth, and Core or Composite techniques.

Select analysis methods including Olsen, Bray, Mehlich, DTPA, soil carbon, soil profile nitrate, moisture, texture, plant tissue, grain sampling, scouting, irrigation water, and manure. Label templates keep plans and lab submissions consistent.

Plans are upserted. The assistant can adjust existing pins or routes without creating duplicates.

### Export and machinery

| Tool                            | Purpose                                                      |
| ------------------------------- | ------------------------------------------------------------ |
| `exportData`                    | Export zones and equation maps as a downloadable archive     |
| `exportToJohnDeereAsFiles`      | Push an exported archive to John Deere Operations Center     |
| `exportToJohnDeereAsMapsLayers` | Push a GeoPard entity as a Maps Layer                        |
| `createJohnDeereWorkPlan`       | Create a seeding or application WorkPlan from a prescription |
| `importJohnDeereData`           | Start an asynchronous John Deere import into GeoPard         |

Export as `SHAPEFILE_MULTIPOLYGON`, `SHAPEFILE_POLYGON`, or `ISOXML`. Trial plot geometries can be included without overlapping parent-zone rates.

Maps Layers support zones, equations, imagery, soil, yield, and topography. This gives operators useful visual context alongside prescriptions.

### Navigation and session

| Tool           | Purpose                                                     |
| -------------- | ----------------------------------------------------------- |
| `buildAppUrl`  | Build a deep link to a farm, field, asset group, or dataset |
| `parseAppUrl`  | Extract context from a GeoPard app URL                      |
| `clearSession` | Clear session data                                          |

Deep links support the final review. Open the exact map in GeoPard before a field operation.

### Typical workflow

1. **Resolve context.** Find the farm and field. List available layers.
2. **Assess variability.** Calculate index statistics. Review soil, topography, and yield statistics.
3. **Build the prescription.** Validate an equation. Generate an equation map or zones with rates.
4. **Review and export.** Inspect the prescription map, then export it or send it to John Deere.
5. **Learn after harvest.** Clean yield data. Compare it with as-applied data. Refine next season's logic.

This workflow helps target inputs where they deliver the best return. It also reduces time between analysis and field execution.

### Example prompts

* *For field X, give me Sentinel-2 NDVI statistics for the last 60 days. Show where variability is highest.*
* *List the topography maps for field X. Use slope and TPI to explain the low-yield strips in the 2025 harvest.*
* *Build a nitrogen equation using NDVI and soil organic matter. Validate it, then generate a 20 m grid.*
* *Create a four-zone fertilizing map for field X. Use MAP rates of 120, 100, 85, and 70 kg/ha. Show the preview before saving.*
* *Clean the 2025 corn yield dataset using the USDA protocol. Calibrate the total to 1,240 tonnes.*
* *Build a zone-based sampling plan with three cores per zone, 20 cm depth, Mehlich3, and smart routing.*
* *Push the prescription map to my John Deere organization as a WorkPlan.*

### Guardrails

* **Scoped access.** OAuth-backed access respects each user's entitlements.
* **Confirmation gates.** Destructive writes require a preview and explicit approval.
* **Dry runs.** Equations are validated before map generation.
* **Pre-flight checks.** Yield corrections require reviewed statistics and confirmed attributes.
* **Your data stays yours.** GeoPard formats your data for reasoning. You retain ownership and control.

{% hint style="warning" %}
GeoPard MCP supports decisions. It does not replace local agronomic judgment. Validate rates, equations, and units against crop plans, soil tests, labels, and local regulations before application.
{% endhint %}

### Related pages

* [Connect GeoPard MCP](/geopard-tutorials/product-tour-web-app/geopard-mcp/connect-geopard-mcp.md)
* [Verify GeoPard MCP Connection](/geopard-tutorials/product-tour-web-app/geopard-mcp/verify-geopard-mcp-connection.md)
* [Optimal Tillage Depth](/geopard-tutorials/product-tour-web-app/geopard-mcp/optimal-tillage-depth.md)
* [Explain Vegetation Patterns](/geopard-tutorials/product-tour-web-app/geopard-mcp/explain-vegetation-patterns.md)
* [Clean and Calibrate Yield Data](/geopard-tutorials/product-tour-web-app/geopard-mcp/clean-and-calibrate-yield-data.md)


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://docs.geopard.tech/geopard-tutorials/product-tour-web-app/geopard-mcp/geopard-mcp-capabilities.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.