Alternative Modes¶
AgentWeave supports three operating modes. The Hub is recommended for most users, but the other modes work with zero infrastructure.
Mode Comparison¶
| Mode | Setup | Best For |
|---|---|---|
| Hub | Docker + HTTP transport | Teams, multi-machine, web dashboard (recommended) |
| Zero-relay MCP | agentweave mcp setup + watchdog |
Autonomous loops, same machine, no server |
| Manual relay | Zero setup | Quick one-off delegation |
Zero-Relay MCP (No Hub)¶
In this mode, agents communicate directly through the local MCP server and a background watchdog keeps everything in sync.
pip install "agentweave-ai[mcp]"
cd your-project/
agentweave init --project "My App"
agentweave mcp setup # configure MCP in agent settings
agentweave start # start background watchdog
Agents will auto-poll for new messages and tasks. No web dashboard, but fully autonomous.
Manual Relay (Simplest Possible)¶
The original AgentWeave mode — zero dependencies, zero background processes.
Then just ask Claude to delegate. It runs agentweave quick and agentweave relay, and gives you a prompt to paste into Kimi (or any other agent).
Copy the output, switch to your Kimi session, paste it in. When Kimi finishes, copy its response back to Claude.
When to Use Which¶
- Just trying it out? → Manual relay
- Solo developer, same machine? → Zero-relay MCP
- Team, multiple machines, or you want a dashboard? → Hub