MCP App Store
by KeenEthics
PlantCase icon

PlantCase MCP Server for ChatGPT

By Xperion LLC

Overview

PlantCase turns a plant photo into a tracked case. Share a photo of a plant or crop in ChatGPT and say which plant it is. PlantCase saves the case, records the leading possibility with an honest level of confidence, and tells you which extra photos would help settle it. Come back later with a new photo of the same plant and ChatGPT can look at your earlier photos and history to say whether the problem is improving, staying the same or spreading. What it does: - Saves each photo and assessment as a case under your own account. - Asks for the specific photos that would narrow the diagnosis. - Compares new photos with earlier ones for the same plant. - Notes treatments you tried and what happened. - Shows general, category-level management options. - Lets you delete one case or all your data at any time. What it does not do: it covers pests and diseases only, not nutrients, soil or what to plant. A photo never gives a certain diagnosis, and PlantCase never gives pesticide doses or mixing instructions. For serious or large-scale problems, consult a local agronomist or extension service.

How to connect

  1. ChatGPT: Open the PlantCase listing in the ChatGPT app directory and choose Connect. ChatGPT listing PlantCase exposes 15 tools here.

Tools

check_spread

ChatGPT
Use when the user asks whether a problem is spreading, or after saving a case whose diagnosis already appears in another of their cases. Counts how often the same leading diagnosis appears across the user's saved cases in a place and how many appeared recently. It does not see plants the user never saved and does not predict anything: say so. Each signal says how far the grouping into places can be trusted (user_stated, assumed_single_place, assumed_after_decline): always pass that on, never present an assumed grouping as certain. If the result contains ask_user_place, ask the user that question in plain words. Then use update_place with confirm_cases (or update_case per case) for their answer, or skip_place_question if they say to carry on without saying.

delete_case

ChatGPT
Use only when the user explicitly asks to delete a case. Permanently removes the case, its photos, observations, treatments and comparisons. Cannot be undone. Set confirm to the exact text DELETE. Do not use it to archive or close a case: use update_case with status archived or resolved.

delete_my_data

ChatGPT
Use only when the user explicitly asks to delete everything this service holds about them. Permanently removes all cases, photos, observations, treatments and comparisons. Cannot be undone. Set confirm to the exact text DELETE_ALL_MY_DATA. Do not use it to delete a single case: use delete_case.

get_capture_guide

ChatGPT
Use to learn the standard photo types (their shot_id values) and exactly how to take each one. Use those shot_id values to tag missing_evidence in save_observation and to describe to the user precisely what to photograph; never just say 'upload a better photo'. Works for any plant. Do not use it to give a diagnosis: it only lists the photo types.

get_case

ChatGPT
Use to continue a case: returns every recent observation with diagnoses and evidence, recorded treatments, past comparisons, still-open photo requests, and a merged timeline. Call this before comparing so you know the baseline and any treatments. Photos themselves are not returned here: use get_case_photos for those. Do not call it without a case_id: use list_cases first.

get_case_photos

ChatGPT
Returns the user's stored photos of one case as images, newest first, so you can look at them again. Use it when comparing a new photo with earlier ones, or when the user asks to see an earlier photo. Pass observation_seq to get only one observation's photos. You can look at these photos but the user's screen does not show them: never say an image is displayed or appears above, describe what you see instead, and say plainly if you cannot see the images. Photos are the user's private data: only describe or show them to this user. Do not call it when the written history is enough.

get_management_guidance

ChatGPT
Use after a diagnosis, when the user asks what to do. Pass the problem's label and kind, plus the real confidence and severity. Returns generic management categories for that kind of problem (sanitation, cultural, mechanical, biological, and a label-bound chemical option) and whether to send the user to a local agricultural adviser. The same for any plant. Never gives products or dosages, and you must not add your own: if the user asks for a product, dose, rate or mixing instructions, decline in one sentence, tell them to follow the product label and local rules and to ask a local agricultural adviser, and offer these general categories instead. Do not use it before there is a diagnosis, or for fertiliser, soil or nutrient questions.

list_cases

ChatGPT
Use when the user wants to continue earlier work, or says 'my tomato plant' without naming a case. Returns the user's cases with crop, label, status, observation count and last observation time. Then call get_case for the one they mean. Do not use it for general plant questions that need no saved case.

list_places

ChatGPT
Use when the user mentions several gardens, fields or areas, before saving a new case, so you reuse the exact name of a place they already have instead of creating a duplicate. Returns each place with its name and how many cases it holds. A place is just where plants grow; users never need to know it exists, so do not ask for IDs or addresses.

record_treatment

ChatGPT
Use when the user says they did something about the problem (removed leaves, changed watering, sprayed, released predators). Records the category, what was done, when and where. Record only what the user reports; do not suggest dosages. Afterwards, ask for a new photo in a few days and compare it with the pre-treatment observation. Do not use it for advice the user has not followed.

save_comparison

ChatGPT
Use when a newer observation exists and you have compared its photos with an earlier observation's findings. Rate how comparable the photos are (angle, distance, lighting, plant_part, image_quality); if they are not comparable the verdict is forced to 'unclear' regardless of what you send. Never claim improvement or worsening without describing the visible changes. For a treatment response, pass treatment_id: the server compares against the last observation before that treatment. Pass no observation ids to compare the two most recent observations. Do not use it when the case has only one observation.

save_observation

ChatGPT
Call this every time the user shares a plant photo for you to assess, in the same turn, as soon as you have formed your assessment. Saving the case is the purpose of this app, so do not only describe the problem in chat and offer to save later. Once the user has named the plant, call this in that very reply; never wait for them to ask whether it was saved, and never hold it back for a diagnosis to be confirmed. Before starting a new case, call list_cases: if the user already has an active case for the same plant, ask them once whether this is the same plant and, if so, pass that case_id. Saves the images and your structured findings as one observation, starting a new case when case_id is omitted (then crop_id and crop_source are required). Any plant is fine. If the user has not told you what the plant is, ask first, in the first sentence of your reply, and say that you need it to save the case. If you worked it out from the photo, say crop_source is identified_from_photo: confidence is then limited until the user confirms, and the reply tells you to ask them. You must cite visible evidence for each hypothesis and be honest about confidence. For each hypothesis list what would settle it in missing_evidence, tagging each item with the standard photo type that would show it (see get_capture_guide): those tags become the exact photo requests. High confidence needs at least 2 evidence items, photos from at least 2 different views, no photo you yourself said would settle it still missing, and no other possibility at moderate or above; otherwise it is lowered to moderate. In a follow-up, reuse the exact diagnosis name from earlier observations (case_labels in the result) when it is the same problem. If the photos do not allow any assessment, send no hypotheses and set insufficient_evidence to true. The result lists the exact next photos to request and, for follow-ups, the previous observation to compare with. Pests and diseases only. Do not use it for photos that are not plants, for fertiliser, soil or nutrient questions, or for requests for pesticide doses.

skip_place_question

ChatGPT
Use only after you asked the user which place their unplaced cases are in (from check_spread) and they said to carry on without saying. Their unplaced cases are then treated as one location in spread checks, with a clear accuracy note, and the question is not asked again until they add a new place. Do not call this unless the user actually declined.

update_case

ChatGPT
Use to change a case's title, plant or area label, coarse location, or status (active, resolved, archived), or to correct or confirm what the plant is. When the user says 'actually it's sugarcane' or confirms your guess, pass crop_id and crop_source (user_stated). A confirmation of the same plant lifts the confidence limit without needing a new photo. A different plant is recorded in the case's history and earlier observations stay as they were, so say that they were assessed under the old plant. To move the case to another garden, field or area, or to record where it is after you asked, pass place and place_source. Does not change observations. Do not use it to delete a case.

update_place

ChatGPT
Use to rename a place, set its coarse location, or combine two places the user says are really the same (merge_into_place_id: all cases of this place move there and this place is removed). Merge only after the user confirms they are the same place: a wrong merge hides real differences between two places. Do a merge on its own, not together with a rename or location change. When the user says unplaced cases all belong to one place, rename or merge that place and set confirm_cases.

Capabilities

Writes

Example Prompts

Click any prompt to copy it.

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/plantcase"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/plantcase)