We’re opening the ZarSage AI private beta today, on macOS first, with Windows and Linux to follow. If you’re an agronomist, a commercial farmer, or an extension officer, and your fields, soil data, and weather history currently live in three different places, this is for you.

You can apply at zarsage.ai/#early-access. Spots are limited.

Why we built this

A single field has soil lab results going back years. It has decades of weather history pinned to its exact coordinates. It has satellite imagery, crop plans, spray records, and a pile of lab reports.

On most farms and advisory practices, the workflow for turning that into a decision looks like: open Excel, cross-reference a PDF, check a weather app, guess.

We think professionals deserve a tool that pulls those inputs into one place and lets AI surface the relevant patterns, without asking you to hand your data over to somebody else’s cloud.

The design choices

Three things shaped the product:

It’s a desktop app, not a SaaS. ZarSage AI runs on your machine (Tauri 2, native binary). Your fields, soil samples, tasks, and cached weather history live on your local disk, SQLite for records, Parquet + DuckDB for the weather archive. You can browse all of it with no internet.

AI is specialized by domain, not generic. When you ask a question, the app routes it to one of four focused contexts, field analysis, operations planning, soil interpretation, or crop research, instead of sending it to a generic chatbot. Each context is tuned for its job.

AI is included with your license. Activate a license and AI calls proxy through our managed gateway, no setup, no separate AI billing, no third-party data sharing. Token usage is metered and visible inside the app at all times.

What’s in the beta

This is what ships on day one, running on macOS:

Field mapping on satellite imagery. Draw boundaries on MapLibre satellite tiles, auto-calculate area in hectares, search locations, and cache tiles for offline viewing. Saved fields render as GeoJSON overlays.
Soil sample vault. Enter lab results through a guided wizard (depth, analytes, methods, coordinates). Samples attach to specific fields and persist locally. Heads-up: it’s manual entry today, no PDF parsing yet.
Historical weather archive. Decades of weather history pinned to each of your field coordinates. Browse historical patterns, compare seasons, and pull hourly forecasts for planning.
Decision dashboards. Five weather widgets on the dashboard: growing degree day tracker, spray window evaluator, water balance, risk radar (frost/heat/rain), and an hourly weather timeline, all calculated from the forecast for your actual fields.
Task calendar with weather scoring. Calendar and Gantt views of your task list. Recommendations get scored against the forecast, spray windows, GDD thresholds, frost risk, so high-stakes tasks land on the right day instead of a convenient one.
Startup briefings. When you open the app, it surfaces what changed: new weather risks, overdue tasks, fields needing attention. No digging through menus to find what needs your eyes today.
Google Tasks sync. OAuth integration to pull in tasks you already track elsewhere.
Calendar export. Any task list can export to .ics so it lands in whatever calendar your team uses.
Specialized AI contexts. Ask field-level questions, ops-planning questions, soil questions, or crop research questions, the app routes each to a domain-specific context with its own knowledge and guardrails. High-stakes recommendations are flagged for your review.

What’s NOT in the beta yet

We’d rather be direct about the gaps than surprise you later:

Windows and Linux builds are close, the CI produces signed installers, but we’re not releasing them to users in this beta. Apply anyway if you’re on Windows or Linux; we’ll reach out when your build is ready.
PDF/CSV lab import. Soil samples are manual entry today. Bulk import is planned, not shipped.
Offline AI. Browsing your data, looking at the dashboard widgets, reading cached weather, all works offline. But AI analysis, weather forecasts, and license checks need a connection.
Mobile. There’s no iOS or Android app. The workflow is desktop-first on purpose.
Shared/team workspaces. It’s single-user today. Multi-user advisory workflows are not built yet.

Who we’re looking for

The first cohort is deliberately small. We want to talk to everyone, not just count signups.

Good fits:

Agronomists and advisors working with multiple farms and tired of switching between tools
Commercial farmers managing several fields who want planning tied to their actual soil and weather
Extension officers who need evidence-based guidance for the growers they serve
macOS users willing to hit bugs, tell us about them, and shape what this becomes

If you’re in, expect emails from a real person, not a support queue.

How to apply

Apply for the private beta, leave your email and a line about what you’re growing or advising on. We’ll reach out when a seat opens.

Questions? Write to hello@zarsage.ai.

Thanks for being early.

, The ZarSage team

The ZarSage AI private beta is open