AgriPilot

AgriPilot is an AI farm operations agent that helps smallholder farmers plan daily work, adapt to weather and feed shortages, and seek approval before costly actions.

Comment

AgriPilot

AgriPilot is an AI-powered farm operations agent for smallholder farmers. It helps turn changing farm conditions into a clear, prioritized daily plan and keeps the farmer in control when a decision involves cost or risk.

The problem

Smallholder farmers often manage crops, poultry, weather, inventory and daily tasks at the same time. A normal farm plan can become outdated within hours when heavy rain arrives, feed runs low, or urgent work appears. Replanning is usually manual, repetitive and easy to miss.

The solution

AgriPilot observes the farm state, generates a prioritized plan, reacts to changing conditions, explains what changed, and requests human approval before costly or higher-risk actions.

The workflow is:

Observe → Plan → Monitor → Replan → Explain → Request Approval

Key features

  • Generates a prioritized daily farm plan
  • Tracks crop and poultry operations
  • Reacts to weather changes such as heavy rain
  • Replans tasks when conditions change
  • Detects low poultry-feed inventory
  • Creates purchase approval requests
  • Keeps payments and costly actions behind a human approval gate
  • Lets farmers mark tasks Done, Delayed or Skipped
  • Maintains a decision trail showing what changed and why

Demo scenario

The live demo uses an AgriPilot demo farm in Gweru, Zimbabwe.

  1. Generate today's farm plan.
  2. Simulate heavy rain.
  3. AgriPilot pauses irrigation and raises drainage-related work in priority.
  4. Simulate low poultry feed.
  5. AgriPilot replans again and proposes a 30 kg feed purchase.
  6. The farmer approves or rejects the recommendation.
  7. The decision trail records the changes.

Human in the loop

AgriPilot is designed to automate routine operational decisions while preserving farmer control. Costly or higher-risk actions require explicit approval. Approval records consent only; payment is never automatic.

AI and agent architecture

AgriPilot integrates the Strands Agents SDK with Amazon Bedrock for agent reasoning and orchestration. Custom farm tools expose farm profile, weather, inventory, task state, replanning and approval workflows to the agent.

For resilience and demo continuity, the prototype also includes a deterministic local planning fallback so the farm workflow can continue if external model access is unavailable or rate-limited.

Technology

  • Strands Agents SDK
  • Amazon Bedrock
  • Python
  • FastAPI
  • React
  • TypeScript
  • Vite
  • Firebase Hosting
  • Render

Deployment

The frontend is deployed publicly on Firebase Hosting and the FastAPI backend is deployed on Render.

AgriPilot is entered in the Professional Agents track because it is designed to make farmers more effective at the work they already do by reducing repetitive planning and replanning overhead.

Built With

Share this project:

Updates