Global Nature Watch (GNW) Horizon · concept

Build your own stack

This concept is different than the other two: it's not a Horizon experience, but three options for reaching GNW's data and AI from other applications and tools.

The orange notes are things to think about as you review each section. The survey you will complete after reviewing asks related questions.
Section 1 · the three options
Three ways to use GNW in other applications and tools Each card below is its own path, not three parts of one flow. Review each and consider the questions at the end.
OPTION A
Get GNW intelligence into an AI assistant you already use. Ask questions about deforestation, land use, or alerts directly inside an AI assistant you already use — without building anything new.
Who this is for Someone who wants conversational access through a tool they already use and didn't have to build.
Technical fluency · light
What it could look like in practice
Your AI assistant
ChatGPT Claude a chatbot on your own site
concessions_2025.shp · shared earlier
How much forest did the Congo Basin lose last year, and what's driving it?
calling global-nature-watch
The Congo Basin lost roughly 580,000 ha of tree cover over the past year — concentrated in the Republic of Congo and northern DRC, mostly from small-scale agriculture. Most of this falls outside your licensed concession boundaries, based on the shapefile you shared earlier. Source: GNW · updated daily
OPTION B
Build something new with GNW using your own dev tools. Point your own coding agent or engineering team at GNW's data and AI to build a custom tool, workflow, or integration.
Who this is for A developer or technical team building their own AI-powered tool — not just pulling GNW's geospatial data, but the tools that let your AI reason over it.
Technical fluency · moderate
What it could look like in practice
Your coding agent writes this
field_report.py · your repository
1gnw = Client(api_key) 2alerts = gnw.glad_alerts(aoi=reserve_boundary, 3  period="last_90_days") 4report = gnw.summarize(alerts, group_by="driver")
>3 new alerts found — report generated
What your team ends up with
Field verification · Q3
3 sites flagged for a ground check on the next patrol round
Block C-741 ha · agriculture Block A-212 ha · roads Block F-118 ha · unclear
A field-verification report your team compiles for local authorities each quarter.
OPTION C
Pull GNW's raw data into my own maps, dashboards, or analysis. Query GNW's data directly — for a map, a report, or your own analysis pipeline — with no AI layer involved at all.
Who this is for Anyone integrating GNW's data into a system they already run — a dashboard, a public map, an alert feed, an analysis pipeline. Unlike Option B, this is data and analytics only, no AI.
Technical fluency · high
What it could look like in practice
GNW publishes this
Docs / Data API / Tree cover loss GET /v1/analytics/tree-cover-loss
Parameters aoi — GeoJSON, or the ID of an area you saved start_year, end_year units — ha or km²
Response
{ "aoi_id": "rbt-c7",
  "loss_ha": 410.2,
  "year": 2024 }
Get an API key
You build this
Your internal dashboard
1,240 ha tree cover loss across our 14 concessions, 2024
Data: Global Nature Watch API · pulled nightly by our own job
Which of these three matches something you would actually want to do? What would you use it for, and what is missing from the set?
If you could bring GNW into wherever you already work, what would that unlock? What would you be able to do that you can't do right now?
Section 2 · how this is powered
For technical reviewers All three options run on the same stack Here's what sits underneath the three options above. It is context for anyone who would build against this — skip it if that isn't your area.
02
MCP SERVICES & AI SKILLS Allows your coding harness or AI agents to easily access GNW's data, including both the geospatial data and instructions on how to interpret it and convert it into different formats.
built on
01
DATA API Analytics and map layers, versioned.
For technical reviewers: does this look like something your team could build against? What would GNW need to publish or guarantee before you'd commit to it?
What this does not do
GNW publishes the interface and the reference code. That is the whole of it — GNW doesn't verify the data you combine with it.
No hosting. Whatever you build runs on your infrastructure. GNW doesn't deploy, monitor, or keep it online.
No accuracy guarantee. GNW doesn't verify the data you combine with its own, and can't confirm the accuracy of results built that way.
What is your overall reaction to this concept? What questions do you have?