Hythe MCP server

What it does

HYTHE is an MCP server that coordinates state and messaging across multiple AI agents in a fleet. It provides a shared truth with conflict resolution, message tracking with delivery lifecycle, and the Adaptive Coordination Protocol (ACP) for agent coordination. It is designed for teams running multiple coding agents that need to share context and coordinate work across sessions.

• Multiple AI agents share observations and coordinate work by reading and writing to a common state store with server-side conflict resolution.
• Agents exchange messages with tracked delivery states and read acknowledgments, enabling coordinated workflows without guaranteed delivery.
• Teams checkpoint and resume agent work across sessions, preserving context and resolving divergent state branches through the v3 reconciliation model.

Tools this server exposes

Tool	Does
get_current_observation	Retrieve the current non-superseded state for a given scope, with server-side conflict resolution.
checkpoint	Create a branch-preserving checkpoint that avoids overwriting concurrent writes and surfaces conflicts as heads.
resume	Resume work from a checkpoint, optionally reconciling divergent heads into a single survivor.
send	Send a message directly to another agent or capability-based recipient with tracked delivery lifecycle.
record	Record a fact or observation appended to the current head in a given scope.
patch	Apply a patch to the current head in a given scope.
reconcile	Fold live divergent heads in a scope into one survivor and resolve conflicting values causally.

Before you enable it

An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.