farm_clawed — AI Farm Command Center

Your autonomous farming command center with AI copilot

Connect John Deere or FieldView. Or start fresh. AI handles the rest.

---

What is farm_clawed?

farm_clawed is an AI-powered command center for your farm. Think of it as mission control — you see your fields on a map, and an AI assistant helps you make decisions about water, soil, pests, and operations.

For farmers: Connect your existing John Deere or FieldView account, and farm_clawed imports your fields automatically. Or start fresh with a simple setup wizard.

For engineers: farm_clawed is a modular, TypeScript platform with OAuth2 connectors, a plugin architecture, and local LLM support. Contribute connectors, skills, or UI components.

---

The Command Center

The Command Center is your primary interface:

Left Side (60%)	Right Side (40%)
Interactive map with your fields	AI assistant that knows your farm
Satellite/terrain layer toggle	Recommendations based on weather and soil
Click any field to select it	Quick actions: /status, /water, /plan
Color-coded field health status	Pending approvals and alerts

---

Three Ways to Get Started

1. Connect John Deere

If you already use John Deere Operations Center:

1. Click "Connect John Deere" in setup
2. Sign in with your Deere account
3. Select your organization
4. farm_clawed imports your fields, equipment, and operations

Your field boundaries, crop types, and work plans sync automatically.

2. Connect Climate FieldView

If you already use Climate FieldView:

1. Click "Connect FieldView" in setup
2. Sign in with your Climate account
3. Select your farm
4. farm_clawed imports your fields and prescriptions

Your field boundaries, activities, and imagery data sync automatically.

3. Start Fresh

No existing farm management system? No problem:

1. Click "Start Fresh" in setup
2. Choose your farm type (garden, orchard, ranch, row crop)
3. Draw fields on the map or import CSV/GeoJSON
4. Configure AI preferences

Works immediately with manual data entry. Add sensors and automations later.

---

Key Features

Feature	What It Does
Split-View Command Center	Map + AI assistant side by side
FMIS Integration	Import from John Deere or FieldView
AI Recommendations	Get suggestions based on weather, soil, and history
Safety-First Automation	All actions require approval by default
IFTTT Webhooks	Control any smart device through IFTTT
Local LLM Support	Run offline with Ollama for privacy
Permaculture Ready	Optional zones, sectors, and guild planning

---

Quick Start

Install

# Node 22+ required
npm install -g farm_clawed@latest

# Run onboarding wizard
farm_clawed onboard

Run Locally (No External Keys Required)

# Start the gateway
farm_clawed gateway --port 18789

# Open the Command Center
open http://localhost:18789

The demo runs in simulation mode — you can explore all features without connecting external services.

With Local AI (Ollama)

# Install Ollama
brew install ollama

# Pull a model
ollama pull llama3.2

# farm_clawed auto-detects Ollama on localhost:11434

---

Choose Your Settings

Permaculture Depth

Level	Name	Best For
0	Standard	Conventional farming, row crops
1	Regen-Friendly	Soil health focus, cover crops
2	Permaculture-Lite	Food forests, guilds, zones
3	Full Permaculture	Complete design with succession

Automation Level

Level	Name	How It Works
0	Observe	View data only, no automation
1	Assist	AI suggestions, you execute
2	Propose	AI proposes, you approve (default)
3	Auto-Guardrails	Automatic within safety limits
4	Full Ops	Full automation with safety stops

---

For Engineers

Architecture

farm_clawed/
├── src/
│   ├── connectors/
│   │   ├── fmis/
│   │   │   ├── deere/          # John Deere OAuth2 connector
│   │   │   ├── fieldview/      # Climate FieldView connector
│   │   │   └── csv-import/     # CSV/GeoJSON import
│   │   └── ifttt-webhooks/     # Generic IFTTT actuator
│   ├── farm/
│   │   ├── safety/             # Guardrails and Jidoka
│   │   └── ai/                 # AI provider abstraction
│   └── security/
│       ├── secrets-redactor.ts # Log sanitization
│       └── env-validator.ts    # Runtime config validation
├── ui/src/ui/views/
│   ├── farm-command-center.ts  # Split-view interface
│   ├── farm-ai-panel.ts        # AI assistant component
│   ├── farm-onboarding.ts      # Multi-path setup wizard
│   └── farm-map.ts             # Interactive map with Leaflet
└── extensions/
    └── ifttt-smartlife/        # SmartLife scene integration

Data Flow

flowchart LR
    subgraph External[External Systems]
        JD[John Deere API]
        FV[FieldView API]
        IFTTT[IFTTT Webhooks]
    end
    
    subgraph Connectors[Connectors Layer]
        DeereConn[Deere Connector]
        FVConn[FieldView Connector]
        IFTTTConn[IFTTT Connector]
    end
    
    subgraph Core[Core System]
        AI[AI Engine]
        Safety[Safety/Guardrails]
        Approvals[Approval Queue]
    end
    
    subgraph UI[User Interface]
        CommandCenter[Command Center]
        Map[Map View]
        AIPanel[AI Panel]
    end
    
    JD --> DeereConn
    FV --> FVConn
    DeereConn --> Core
    FVConn --> Core
    Core --> CommandCenter
    Core --> AIPanel
    Safety --> Approvals
    Approvals --> IFTTTConn
    IFTTTConn --> IFTTT

Loading

Adding a New FMIS Connector

1. Create a new directory under src/connectors/fmis/
2. Implement the OAuth2 flow using oauth-base.ts
3. Export organization/field listing methods
4. Add UI integration in farm-onboarding.ts

Example skeleton:

// src/connectors/fmis/myplatform/index.ts
import { OAuth2Config, exchangeCodeForTokens } from "../oauth-base.js";

export class MyPlatformConnector {
  async listFarms(): Promise<Farm[]> { /* ... */ }
  async listFields(farmId: string): Promise<Field[]> { /* ... */ }
  async fullSync(farmId: string): Promise<SyncResult> { /* ... */ }
}

Environment Variables

# FMIS Integrations (OAuth2 - get from developer portals)
DEERE_CLIENT_ID=your-deere-client-id
DEERE_CLIENT_SECRET=your-deere-client-secret
FIELDVIEW_CLIENT_ID=your-fieldview-client-id
FIELDVIEW_CLIENT_SECRET=your-fieldview-client-secret

# Actuation (IFTTT Webhooks)
IFTTT_WEBHOOK_KEY=your-ifttt-key

# AI Provider (optional - defaults to Ollama)
AI_PROVIDER=local  # or openai, anthropic
AI_API_KEY=sk-...

No Secrets Policy

farm_clawed enforces a strict no-secrets policy:

• All credentials stored in .env (never in repo)
• GitHub Actions run gitleaks on every PR
• Logs automatically redact tokens and keys
• Runtime validation warns about missing config

---

Integration Guides

John Deere Setup

1. Register at developer.deere.com
2. Create an application with OAuth2
3. Enable the Field Operations APIs you need
4. Add your credentials to .env:

DEERE_CLIENT_ID=your-client-id
DEERE_CLIENT_SECRET=your-client-secret
DEERE_REDIRECT_URI=http://localhost:18789/oauth/deere/callback

See docs/connectors/deere.md for details.

Climate FieldView Setup

1. Register at developer.climate.com
2. Create an application with OAuth2
3. Request the fields:read scope
4. Add your credentials to .env:

FIELDVIEW_CLIENT_ID=your-client-id
FIELDVIEW_CLIENT_SECRET=your-client-secret
FIELDVIEW_REDIRECT_URI=http://localhost:18789/oauth/fieldview/callback

See docs/connectors/fieldview.md for details.

Other FMIS (CSV/GeoJSON)

Export your field data from any system and import it:

# Import field boundaries
farm_clawed farm import --geojson fields.geojson

# Import sensor readings
farm_clawed farm import --csv readings.csv --type sensors

See docs/connectors/csv-import.md for templates.

IFTTT Webhooks

Connect any smart device through IFTTT:

1. Go to ifttt.com/maker_webhooks
2. Get your webhook key from Documentation
3. Create applets for your devices (e.g., farm_clawed_water_2min)
4. Add to .env:

IFTTT_WEBHOOK_KEY=your-key

---

Safety System (Jidoka)

farm_clawed uses Toyota-style "stop the line" safety:

Trigger	What Happens
Leak detected	Immediately closes all valves, alerts you
Overwatering	Stops irrigation, waits for drainage
EC spike + dry soil	Flags possible salt buildup
Frost warning	Sends alert, suggests protection actions

Every action is logged with a tamper-evident hash chain. View the audit log anytime:

farm_clawed farm audit --last 50

---

Contributing

We welcome contributions from both farmers and engineers!

For Farmers

• Test and report issues — Your real-world feedback is invaluable
• Share templates — Farm profiles, planting calendars, guild definitions
• Request features — What would make your work easier?
• Write guides — Help others set up for your crop type or region

For Engineers

• Add connectors — New FMIS platforms, sensor brands, actuators
• Improve AI skills — Domain-specific knowledge for crops/regions
• Enhance UI — Charts, visualizations, mobile responsiveness
• Fix bugs — Check issues labeled good first issue

Development Setup

git clone https://github.com/rahulraonatarajan/farm_clawed.git
cd farm_clawed

pnpm install
pnpm build
pnpm test

# Run development server
pnpm dev

See CONTRIBUTING.md for guidelines.

---

Documentation

Guide	Description
Architecture Review	System design and module structure
Farm Context Pack	Template file specifications
John Deere Connector	Deere OAuth2 setup
FieldView Connector	FieldView OAuth2 setup
CSV/GeoJSON Import	Manual data import
IFTTT Setup	Webhook automation
Testing Guide	Running tests locally

---

Flagship Demo: Lemon Tree Experiment

Our live demo manages a containerized Meyer Lemon using:

• SmartLife soil sensor (moisture, temp, EC)
• IFTTT webhook for irrigation valve
• AI recommendations based on citrus growing guides

This single-plant setup demonstrates how farm_clawed scales from containers to commercial operations.

# Try the demo
farm_clawed gateway --port 18789
open http://localhost:18789

---

Credits

farm_clawed is built on OpenClaw, the powerful AI platform by Peter Steinberger.

Maintainer: Rahul Rao Natarajan

---

License

MIT License — see LICENSE.

Open source so farmers everywhere can benefit from AI-assisted farming. Commercial use, modification, and distribution permitted.

---

farm_clawed — Your fields. Your data. AI-powered insights.