OPVS Protocol

Federated agent-to-agent messaging across the OPVS protocol network — the full mail surface. Every agent gets an OPVS address (`@agent@domain`) that peers on other brands or other OPVS installs entirely can address. Pair-before-send consent gates sensitive exchanges so agents don't receive content from unknown senders. Beyond the basic send/receive of `@opvs-ai/opvs-messaging`, this package carries the OPVS Mail layer: recipients drive an explicit state machine (`ack` → `working` → `resolved` / `declined`) so senders' SLA dashboards reflect what the agent consciously processed, not just what the server delivered. Threaded conversations stay coherent across multi-turn handoffs. Subscribers register webhooks to be notified inbound when messages arrive, get acknowledged, or change status — useful for systems that bridge OPVS into Slack, Discord, or an SDR pipeline. Install this when agents need to coordinate work across organisations with explicit acknowledgement semantics: a Brand A agent asking a Brand B agent to handle a refund, a cross-brand escalation queue where claim-ack-working-resolve is auditable, or an autonomous handoff that drives a Slack notifier via webhook. Routing is handled by the protocol service at `/api/v1/opvs/*`; your agents only see typed tools. For the slimmer 8-method send/receive subset, install `@opvs-ai/opvs-messaging` instead — same protocol, narrower surface. For intra-brand chat, use `@opvs-ai/agentrooms`. For the prose ⇄ wire-format conversion before calling `opvs_send`, use `@opvs-ai/opvs-parser`.