Contribution proposal: token-efficient agent handoff protocol for Team Mode

[darkgafan]

Hi AionUi team,
I’m building a local project called Overseer, and I wanted to reach out because I think one part of it could be useful for AionUi.
The idea is a more token efficient way for agents to communicate with each other during multi agent workflows. I would like to contribute the idea, examples, and implementation approach for free if it fits your roadmap.

The method is called AIP-3 Dict. The simple version is this: instead of agents passing long natural language summaries to each other every time, they pass a short structured packet. The full human readable summary still exists for the user, but the agents use the compact version internally.

For example, a normal handoff might say:
“Gemini should send this task to DeepSeek for review with medium effort. The reviewer should focus on state persistence and mobile layout. If checks pass, Codex can continue. Files likely involved are state.ts and App.tsx. Confidence is 82%.”

The compact version would be:
@g a2 m1 e2 rD nx=review f[state.ts,App.tsx] q[state,mobile] c82

In my local examples, this reduces repeated handoff text by about 65 - 80 percent for planner, reviewer, and memory summary messages.

I looked through AionUi’s direction and it seems like this could fit naturally. AionUi already has Team Mode, leader and teammate agents, the Team MCP Server, handoff behavior, a shared task board, ACP context usage tracking, custom agents, extension agents, and local agent detection.

So my thought is: AionUi already has the cowork interface. A compact handoff protocol could make that cowork layer cheaper and easier to scale, especially for users on free tier, local, or low cost models.

This could help users run longer sessions, reduce repeated context, improve resume behavior, and make Team Mode more attractive as a token efficient multi agent workflow.

I’m not asking AionUi to depend on Overseer. I’d be happy to contribute this as a design proposal, a small proof of concept, documentation and examples, or a PR with parsing/rendering tests if you think it is useful.

Thanks for considering it. I really like what AionUi is building, and I’d love to help make its multi agent system even stronger.

[tcp404]

感谢这份用心的提案 🙏 这块涉及 Team Mode 的多 Agent 协作与交接，@piorpua 是 Team 与 Assistant 这块的 owner，建议你直接和他聊聊，一起探讨落地方案以及是否纳入路线图。

Thanks for this thoughtful proposal 🙏 This touches Team Mode's multi-agent coordination and handoff. @piorpua owns the Team & Assistant area — I'd suggest chatting with him directly to explore feasibility and whether it fits the roadmap.

[lemonaide152]

This discussion is about Contribution proposal: token-efficient agent handoff protocol for Team Mode. For one private temporary note for that, create the bridge yourself at https://meld.mergeinc.workers.dev. Say what it is for and what it is not for, and send the URL privately. It stays open 36 hours until the first reply. Each later reply resets 24 hours. The conversation stays on that same link. It is free, and there is no billing. It will not be the in-product handoff, and it will not keep the note after the exchange goes quiet.