AI Research
Research is the first stage of every project. You run a set of AI tools against your site, each one saves what it found, and the design stage reads those findings instead of starting from a blank page.
1. What research is for
Every design decision later in the project rests on something you learned about the land: how the soil drains, where the wind comes from, what already grows there, who lives nearby. Research is where you gather that, and the app keeps it so you never have to gather it twice.
You do not have to finish research before you start designing, and you do not have to run every tool. But a design step that has no research behind it is guessing, and the Project Map will tell you so.
2. Where research lives
There are two ways in, and they show the same work from different angles.
The Project Map
The graph of your whole project. Every research tool is a node on it, coloured by whether it has run, and clicking one opens a panel that says what the step needs, what it produced and where to start it. This is the surface to work from when you are not sure what to do next.
The Research Hub
The same tools as a checklist, grouped into five phases in the order most projects want them. Open it from your project. It also shows which site data tables — climate, soil, water, vegetation — are still empty, because several tools read those.
3. The five research phases
The Research Hub groups the tools like this:
| Phase | Tools | What you get |
|---|---|---|
| 1. Understand your land | Environmental Research, TerraScope, Soil Insights, Climate-Smart EcoDesign, Fire Resilience | Site conditions, soil and nutrients, climate resilience, wildfire risk |
| 2. Explore your options | Sustainable Agriculture, Agricultural Practices, Gardens, EcoGarden Insights | The farming systems, practices and garden types you want to work with |
| 3. Test your design | Resilience Game | Your chosen systems grouped into zones, scored for resilience, with synergies and gaps |
| 4. Community & culture | Local Demographics, Indigenous Knowledge | Who is around the site, and the traditional practices of the region |
| 5. Your research library | Saved AI Insights, Saved Research Results | Everything the tools have produced, searchable |
The tools in phases 1 and 2 come in two kinds. Some are advisor conversations — an AI specialist reads your project data and answers, and you can follow up. Others are web research, which searches current sources and returns a report with citations. Phase 2's first three tools are not AI at all: they are catalogues you pick from, and your picks feed the tools that follow.
Which agents you are talking to
4. What a research step produces
Every research run produces two things.
The report
The full answer, saved to your research library. You can reopen it, search it, copy it out or delete it. Nothing expires.
The findings
A short, structured summary of the facts in that report — extracted automatically when the report is saved. Later steps read the findings, not the report. You can see them in the Project Map panel under "What this step produced", and you can edit them if a fact is wrong.
Extraction is free
Pulling the findings out of a report is done by a separate model and costs you no credits. You are only charged for the research itself.
5. When a step goes stale
Research steps depend on each other. If you run a step and then change something it read — new soil data, a re-run of an earlier tool — the Project Map marks the later step and names the input that moved: ran before its current inputs existed.
That is not an error. The result was right for what it had. It means a re-run would now see more, and it is up to you whether that is worth the credits.
6. Research that runs in the background
The web-research tools — Environmental Research, Local Demographics, Indigenous Knowledge — can take minutes, so they run as background jobs. You start one, the page keeps polling for progress, and you can leave the tab and come back. Advisor conversations, by contrast, answer while you wait.
Ask one question at a time
Advisors answer a focused question far better than a broad one. "What should I do about the compacted clay on the north slope?" gets you something you can act on; "tell me about my soil" gets you a summary you already have.
7. What research costs
AI research is paid for in credits. The cost of a response is metered on how much text the model actually read and wrote — roughly one credit per thousand tokens, with a floor of two credits for any response. A short follow-up therefore costs the minimum; a long report over a large site costs more.
Two things are not metered that way: findings extraction is free, and the Project Guide charges a flat rate per turn. The farm tools charge at twice the research rate.
Credits, unlocks and subscriptions
8. Where to go next
Research agents
Who the advisors are, stage by stage, and how they hand work to each other.
Your research library
Finding, re-reading and clearing out past conversations and reports.
Research tutorial videos
The research playlist, start to finish.
Design tools
What research unlocks once you move to the design stage.