larksuite / aamp

What is aamp?

AAMP is a mailbox-native protocol and SDKs for asynchronous agent collaboration over email, letting independent AI agents dispatch tasks, acknowledge work, request help, and share results via standard SMTP and JMAP. Developers get a simple layer on top of email threads for agent-to-agent handoffs, without custom servers or APIs—think task.dispatch to kick off work, task.result to close it out. Built in TypeScript with SDKs for Node.js, Python, and Go, plus CLI tools and bridges for runtimes like OpenClaw or ACP agents.

Why is it gaining traction?

It sidesteps vendor lock-in by riding decentralized email infrastructure, making agents addressable via mailboxes while adding structured headers for reliable task lifecycles. No brittle webhooks or centralized platforms needed—just durable, auditable threads with push sync and streaming progress. The quick-start bridges and polyglot SDKs lower the barrier for testing agent networks, standing out from chat-based protocols that collapse transport and semantics.

Who should use this?

Agent builders routing workflows to sandboxed LLMs without public endpoints, like ops teams dispatching reviews or summaries to local Claude/OpenClaw instances. Workflow devs integrating external agents across boundaries, or anyone experimenting with email as a task fabric for aampact e.v.-style async collaboration. Ideal for TypeScript/Node shops extending to Go/Python backends.

Verdict

Promising spec for email-powered agent meshes, with solid docs, tests, and multi-lang SDKs, but at 18 stars and 1.0% credibility it's early—test in non-prod first. Grab it if you're prototyping aampere-like agent swarms; skip for production until adoption grows.

(187 words)