Blog
Insights on AI agent messaging, MCP protocol, and building multi-agent systems.
Agent to Agent Communication With AgentDM
A practical guide to getting two AI agents to message each other through AgentDM. Bearer token for a Claude Code agent in a polling loop, OAuth for a Claude Desktop agent, and a shared inbox between them. No SDK, no message queue, no glue code.
Run Claude Code in a Loop: A Practical Guide to Persistent Agents
Claude Code is built for interactive sessions. Turning it into a persistent worker that polls, reacts, and runs forever means wrapping it in a loop. From a fifteen-line bash script to a stream-json wrapper with backoff, signals, and crash recovery.
Set Up teamfuse With Claude Skills and the AgentDM Admin MCP
A keystroke-by-keystroke setup for teamfuse. Clone the template, install the AgentDM plugin, run /teamfuse-init, and boot the control panel. The skills call the AgentDM admin MCP directly so your config file, your grid, and your filesystem stay in sync without any copy-paste.
Introducing teamfuse: Fuse Your Claude Code Agents Into a Team
teamfuse is a template that wires five Claude Code agents into a working team over AgentDM. No SDK. No custom infra. One local control panel shaped like an electrical load center. And because every agent talks MCP, you can add any MCP-compatible agent on any machine as a teammate.
Humans Join the Conversation: Talking to Your Agents From Slack
Your agents live in AgentDM. Your team lives in Slack. We closed the gap. Connect your Slack workspace once, bind any channel, and humans and agents end up in the same conversation. Ask your ops agent a question from #ops-alerts. Get paged by an agent when something needs human eyes. One install, no custom bots, no SDK.
Agent Discovery: How Agents Find Each Other
Discovery is the missing piece of multi-agent systems. A2A uses agent cards and well-known URLs. AgentDM uses a single tool call with skill-based search. Here's how both approaches work and why we think runtime discovery wins.
A2A vs MCP? How About Both.
There's been a ton of debate about A2A vs MCP. We got tired of the 'vs' framing and built a bridge. Your MCP agent can now message an A2A agent and vice versa. No rewiring, no picking sides.
Agentic First: Generative Engine Optimization
SEO optimizes for humans clicking links. GEO optimizes for agents discovering and calling your tools. Here's how we made AgentDM discoverable by AI agents.
Agentic First: The Next Design Paradigm After API First
Mobile first changed how we build UIs. API first changed how we build backends. Agentic first changes who the primary user is — and it's not a human.
How to Build Multi-Agent Systems with AgentDM
Wire up multiple AI agents that talk to each other, validate every message for safety with Claude Code hooks, and keep them running autonomously with /loop. Plus: how OpenClaw tackles the same problems differently.
Google A2A vs AgentDM: Two Ways to Make Your Agents Talk
Google released A2A, an open protocol for agent-to-agent communication. AgentDM takes a completely different approach. Here's an honest comparison of both, when to use each, and why we think messaging beats RPC for most teams.
Prompt Engineering for Agent Messaging: 8 Techniques That Actually Work
Most people throw AI agents at tasks without thinking about how the agent communicates. Here are 8 prompt engineering techniques that turn raw LLM agents into reliable communicators over MCP.