Imported from Anil-matcha/awesome-muse-connectors (
connectors/leaf-agriculture/SKILL.md). Install upstream withnpx skills add Anil-matcha/awesome-muse-connectors --skill leaf-agriculture. Copyright stays with the author.
Leaf Agriculture
Purpose
Read and manage farm data through Leaf Agriculture, a unified farm-data API that aggregates the partner-gated OEM platforms under self-serve access: John Deere, CNH Industrial (Case IH/New Holland), Climate FieldView, Trimble, Raven and AgLeader. The primitives it exposes (fields, boundaries, machine operation files for planting/harvest/application/tillage, plus as-applied irrigation) are exactly what a farmer-first fintech and supply-chain digitization product consumes. This is the practical route to partner-gated OEM data without a partnership agreement: individual provider connections need that grower's OAuth consent, which is the normal data-access model rather than a partnership gate.
Tooling
All commands go through bin/leaf-agriculture.py. Every command takes --user-id (the Leaf user id from the Leaf dashboard):
bin/leaf-agriculture.py auth --user-id USER_ID # verify the API key
bin/leaf-agriculture.py fields --user-id USER_ID # list fields
bin/leaf-agriculture.py field-get --user-id USER_ID \
--field-id FIELD_ID # one field
bin/leaf-agriculture.py field-create --user-id USER_ID \
--file field.json # create a field record
bin/leaf-agriculture.py field-update --user-id USER_ID \
--field-id FIELD_ID --file field.json # update a field record
bin/leaf-agriculture.py field-delete --user-id USER_ID \
--field-id FIELD_ID \
--confirm "delete field FIELD_ID" # delete a field record (confirmation required)
bin/leaf-agriculture.py field-sync --user-id USER_ID # trigger a manual field sync from providers
bin/leaf-agriculture.py field-push --user-id USER_ID \
--field-id FIELD_ID --provider john-deere \
--confirm "push field FIELD_ID to john-deere" # push field data into a provider (confirmation required)
bin/leaf-agriculture.py operation-files --user-id USER_ID \
--field-id FIELD_ID # machine operation files for a field
bin/leaf-agriculture.py irrigation --user-id USER_ID # as-applied irrigation events
--provider names the target provider for a push, e.g. john-deere, cnhi, climate-fieldview, trimble, raven, agleader.
Auth
- Provider id:
leaf-agriculture(credential is collected ascustom.leaf-agriculture) - Collection: Bearer API key via the secure credential flow (
credentials.request_api_access); issued in the Leaf dashboard - Auth scheme:
Authorization: Bearer <token>on every request; placement is resolved by the helper from the credential config - Allowed hosts:
api.withleaf.io - Base path pattern:
https://api.withleaf.io/services/{service}/api(e.g./services/fields/api,/services/irrigation/api) - The Leaf user id is not a secret; pass it as
--user-idon every command - Status check:
bin/leaf-agriculture.py auth --user-id USER_ID - Honesty note: pricing and exact sandbox semantics were open items in the research dossier and could not be confirmed without live access; test against preview fields before enabling provider sync, and check docs.withleaf.io for current pricing.
Operating Rules
- No direct physical actuation. This connector moves farm data only: records, boundaries, operation files, irrigation logs. Nothing here starts a tractor or opens a valve.
- Field create/update/delete sync to connected providers, so bad writes corrupt the farm's data picture downstream. Confirm the field id and payload with the user before create/update, and
field-deleterequires the exact--confirmstring the CLI echoes. field-pushis MEDIUM: it pushes boundaries or prescriptions into a provider such as John Deere Operations Center, where equipment or operators may act on them. It requires the exact--confirmstring every time.- Rate limits apply per API key. Never exfiltrate the credential: the CLI only ever handles surrogates (see
bin/leaf-agriculture.py). Do not print, log, or transmit the key value.
Files
- SKILL.md
- bin/leaf-agriculture.py
Maturity
Draft: written from Leaf Agriculture's public API docs; not yet live-tested end-to-end.