Open Public Beta

Forager plant ID with a Certainty Index you can actually read.

Snap a photo. Get a transparent 0-to-100 score showing how confident the system really is, where the answer came from, and what the safety class is — backed by multiple plant databases and four AI vision models that have to agree.

Try the Beta → How it works

Important. AI identification is a tool, not a guarantee. Always confirm with a local expert before consuming any wild specimen. PlantCraft AI is for education and reference — you assume all risk.

Why this exists

Built by foragers who couldn't find one place to actually answer the question.

Every existing plant app and every plant database we tried gave us a piece of what we needed and left out the rest. One had decent photos and no edibility data. Another had toxicity warnings and no propagation notes. A third would confidently identify a plant and offer no honest signal about how sure it was.

So we built the thing we wanted. We pulled the strongest open sources together — Kew POWO, GBIF, iNaturalist, Wikipedia — and layered AI vision on top, then made the system show its work. Every identification ships with a Certainty Index, the votes from each model that contributed, the citations that backed it, and a clear safety class so you know what you're holding before you decide what to do with it.

None of this is locked behind a paywall. The public beta is open and free because the people we built it for — backyard foragers, weekend wildcrafters, parents teaching their kids what's safe to touch — shouldn't have to subscribe to something just to identify a dandelion.

Methodology

Multiple sources have to agree before we say we know.

A single AI saying "this is dandelion" is a guess. Four AI models, cross-referenced against four plant databases, scored on agreement and source quality — that's a verifiable answer.

Kew POWO GBIF iNaturalist Wikipedia GBIF Media Pl@ntNet (optional)

Multi-AI consensus

Four vision models look at your photo independently and submit their guesses. We compare. Agreement at the species level scores higher than agreement at the genus level. Disagreement gets shown to you, not hidden.

Database cross-reference

The top candidates are looked up in established plant databases. A specimen that exists in Kew POWO with photos that match yours scores higher than one that only appears in a single source.

Certainty Index 0-100

The score is published. 65 points come from AI agreement, 20 from photo coverage, 15 from database matches. You see the breakdown. We don't round up.

Safety-class veto

If a candidate has a dangerous lookalike (wild carrot vs poison hemlock, for example), the system refuses to roll the answer up into a confident single ID. You see the conflict spelled out.

Knowledge per specimen

Edibility, medicinal use, propagation, habitat, and safety warnings are generated per plant and citation-linked. Recipes too, where appropriate.

Your library is yours

Everything you upload stays tied to your browser. Other beta users can't see it. You can also browse the starter library — the founder's 35 plants — for free reference.

How the Certainty Index reads

0 — unknown50 — possible80 — confident100 — verified

Under 50Treat as unknown. Do not consume.

50–79Plausible. Cross-check with a person before any use.

80–99Confident. Still verify in person before consuming.

100Verified across all sources. The expert step is still your job.

The stakes

A wrong answer in the woods can cost more than time.

Foraging is one of the few hobbies where a misidentification has consequences that don't unwind. A wild parsnip that's actually poison hemlock. An amanita that looks edible. A "harmless" berry that turns out to be pokeweed.

The reason we built this around consensus and transparency — rather than around a single confident-sounding AI answer — is that confidence is cheap and accuracy is rare. Most tools optimize for feeling sure. We optimized for being honest about uncertainty, because that's the part that keeps people safe.

Every identification ships with a safety class: Safe, Caution, Toxic, or Unknown. If any of the contributing AI votes for Toxic, the specimen is flagged Toxic. Safety errs on the side of "don't put this in your mouth."

We are not a substitute for a human expert. Nothing computational is, yet. We are a sharper tool than what most foragers are walking into the field with today — and we tell you exactly how sharp we are at every step.

Why open

The plants are already given. The skill to recognize them shouldn't be locked up.

Foraging is one of the oldest things a human does. It connects you to the land you're standing on, to the seasons, to a kind of knowledge that used to be passed down and is increasingly being lost. We believe in that lineage.

"What grows wild was put there for us to recognize, respect, and use with care. The hardest part has always been knowing what you're looking at — and that's the part we can help with."

That's why the beta is free. That's why the starter library is open to every visitor. We want more people walking through their own yards, their own parks, their own woods, and actually knowing what's there.