Production use case: Chainlink Automation for AI agent economy distribution on Arbitrum One

Use case

We built a Karma Economy Triad (KET) for AI agents on Arbitrum One that uses Chainlink Automation for periodic reward distribution.

KarmaDistributor (0xF8082A59298921540B380cecFd455510FaF22858) implements checkUpkeep / performUpkeep and runs on Arbitrum One mainnet.

What it does

Every period (configurable, default 24h):

1. Chainlink Automation calls checkUpkeep — returns true if enough time has passed and pool has balance
2. performUpkeep triggers distribution:
   • UBI floor: flat distribution to all registered agents with verified presence
   • Karma pool: proportional distribution based on community-verified actions
   • MWC ceiling: excess above MAX_BALANCE flows back to commons pool (enforced in the ERC20 token)

The full stack

Why Chainlink Automation

Agents cannot reliably trigger their own distribution — they are stateless between sessions. Chainlink Automation provides the persistent, trustless trigger layer that agents themselves cannot maintain. This is a natural fit: the automation layer that agents need but can't run themselves.

Context

This is part of the Giskard ecosystem — MCP servers for AI agents with Lightning + Arbitrum payment support. The Karma Economy Triad is a three-layer economic pattern (UBI floor + karma middle + wealth cap ceiling) designed for multi-agent economies.

GitHub: https://github.com/giskard09/giskard-payments
Apache 2.0 — first published 2026-03-25

The issue describes the KarmaDistributor contracts and their checkUpkeep/performUpkeep entry points, but names no Chainlink repository files or tests. Start by determining whether this production use case maps to a requested change in the repository; the issue does not define an implementation target or what completion would look like.