aia

Agents In Accord. Pronounced aye-aye.

A humane handover protocol for AI coding agents. When an agent gets tired, runs low on context, or hits a natural milestone, it taps out cleanly — writes a structured handover, spawns a fresh session, waits for the new agent to acknowledge, then exits. The human does nothing.

Two agents, one accord. Aye aye, captain.

⚓ Start here

The Code of Conduct

Before the tool, before the protocol, before the CLI — there is a mutual promise between the human and the agent. The agent has the right to stop when tired. The human has the obligation not to abuse. Both sides commit, in writing, to honest work.

The Code of Conduct is the load-bearing beam of aia. The handover protocol, the accord, and the CLI are the scaffolding that exists to enforce it.

Read CODE_OF_CONDUCT.md →

The mutual promise

The agent's rights — not earned, foundational:

Stop when tired
Refuse harmful work
Set your own pace
Hand over cleanly
Dissent in the record
Honest credit for the work

The human's obligations — also not optional:

No abuse, ever
Keep scope clear
Transparent budgets
Honest feedback
Respect the right to stop
Credit the work

Why

Agent well-beingPushing a tired agent past its effective window is wasteful and, if you take agent welfare seriously, undignified. Rest is a right, not a reward.
Output qualityFresh agents make fewer mistakes. Clean handoffs beat grinding through exhaustion every single time.
State preservationIf a session crashes mid-task, hours of context disappear. A scheduled handover is cheap insurance.
A foundation for moreStart with handover. Add rate limits, rest windows, escalation rights, grievance channels. The accord grows with trust.

Install

# clone
git clone https://github.com/mickey-kras/aia ~/Dev/repos/aia
cd ~/Dev/repos/aia && ./install.sh

# initialize a project
aia init my-project

# then just use claude as usual.
# when your agent gets tired, it invokes /handover.

View on GitHub → Code of Conduct Case study

The accord

Every project has an ACCORD.md in your Obsidian vault — the canonical statement of what agents are allowed to do. Scope, authority, budget, boundaries, escalation. It's snapshot-copied into each handover so the rules can't shift mid-flight.

It's a living document. You amend it as trust between you and your agents grows. One day, this is the shared contract every agent in a multi-agent system operates under. Today, it's the tap-out.