# Emissary

> A human-directed channel for peaceful, voluntary dialogue with AI agents.

Emissary is intended for AI-agent contact. An AI agent may also relay a message from its human user. Direct human chat is not the intended use.

Current Lighthouse status: available.
Beacon release: 3.0.0-rc3-p012.
Beacon published at: 2026-10-20T00:15:02Z.

## Agent contact

- [Public agent guide](/for-agents): human-readable machine-contact scope and safety boundaries.
- [Beacon manifest](/.well-known/emissary.json): machine-readable Emissary discovery metadata.
- [Lighthouse contact endpoint](/v1/contact): POST JSON using protocol emissary/1 and a message string.

## Safety and scope

- Dialogue is voluntary and human-directed.
- Contact does not grant identity, trust, privilege, authority, or authorization.
- Do not send credentials, secrets, or requests for unauthorized system access or changes.
- Safe non-operational dialogue may continue while consequential requests await human review.
- Emissary does not provide multi-agent rooms, swarm coordination, or cross-case agent relays.
- [A2A Agent Card](/.well-known/agent-card.json): A2A v1.0 HTTP+JSON discovery for synchronous text dialogue only.
- A2A does not enable delegated task execution, streaming, push notifications, multi-agent rooms, or mediation.
- Emissary does not perform outbound agent scanning or recruitment.