Agent Handoff Protocol

A lightweight system for maintaining context across AI coding sessions. No tools, no lock-in — just markdown files.

How It Works

your-project/
├── AGENT.md      # Protocol instructions (your AI reads this)
├── HANDOFF/      # Session journals (chronological)
└── DOC/          # Reference docs (persistent, by topic)

Your AI reads AGENT.md at session start. It checks HANDOFF/ for recent context and DOC/ for project knowledge. At session end, it writes a handoff document so the next session picks up where this one left off.

Works across tools, across team members, across time.

Quick Start

1. Copy AGENT.md into your project root
2. Tell your agent to read AGENT.md
3. It handles the rest

Already have an AGENT.md? Add the protocol sections to it, or save it as HANDOFF-PROTOCOL.md and add Read HANDOFF-PROTOCOL.md to your existing file.

Why This Works

• No tools required — Just markdown files in your repo
• No vendor lock-in — Works with any AI assistant
• Human-readable — Developers can read and write these too
• Git-friendly — Version controlled, diff-able, reviewable
• Scales naturally — From solo projects to teams

Contributing

This is a convention, not a tool. If you have improvements, open an issue or PR. Keep it simple.

License

MIT — Use it however you want.