agent-mq is a message queue designed for AI programming assistants, with the goal of enabling AI agents to communicate with each other. According to the project materials, it supports agent message passing across sessions and across machines, and can be self-hosted.
In terms of purpose, agent-mq is used to solve the problem that multiple AI programming assistants in the same or different environments are isolated from each other and cannot directly exchange information. Through the message queue, agents running in different sessions or on different machines can send messages to each other, thereby working collaboratively.
In terms of main capabilities, the materials explicitly mention: it provides message queue capabilities for AI programming assistants, allowing AI agents to communicate with each other; it supports integration via MCP, for example by writing to configuration files such as ~/.claude/mcp.json and ~/.cursor/mcp.json; it also supports installation as a Claude Code plugin; it can be used by logging in through one's own server; except for specially noted endpoints, all endpoints require authentication using Authorization: Bearer <token>.
In terms of usage, you can first add agent-mq to the MCP configuration, or install it as a Claude Code plugin; then have the AI agent learn how to use it according to the provided documentation address; then use your own server to complete login. The materials do not provide specific installation commands, configuration field examples, server deployment steps, or token acquisition methods; these contents are not currently provided.
In terms of applicable scenarios, it is suitable for users who need to pass messages between multiple AI programming assistant sessions, want agents on different machines to communicate collaboratively, and have self-hosting requirements.
In terms of limitations, the materials do not explain the supported message formats, throughput capacity, deployment environment requirements, license, whether it is paid, or pricing information; these are not currently provided.
