Feishu Multi-Agent Configuration Guide

Problem

A single Feishu bot often routes all group requests to one agent, making prompts and context grow as roles multiply. feishu-multi-agent-s addresses the setup problem of binding multiple OpenClaw agents to multiple Feishu bot apps: each agent gets its own AppID/AppSecret, several bots appear in the same group, and the agents can still coordinate inside OpenClaw.

How It Works

The skill is mainly a configuration guide for mapping OpenClaw resources to Feishu open platform apps:

• Each agent needs a dedicated Feishu app with the bot capability and WebSocket long connection enabled, subscribing to im.message.receive_v1 without a public callback.
• OpenClaw configuration uses agents.list, channels.feishu.accounts, bindings, and channels.feishu.accounts.default to attach credentials to agents and control access.
• Since Feishu WebSocket does not deliver other bots' messages as events, inter-agent communication relies on OpenClaw primitives such as sessions_send, sessions_list, and sessions_spawn.
• Collaboration policies are written in SOUL.md or AGENTS.md, where prompts decide when to call another agent.

Boundaries

It fits teams already using OpenClaw that want a Feishu group as a multi-agent workspace. For a single-bot, single-agent use case, managing N Feishu apps adds operational overhead. Real division of labor depends on prompt rules and explicit sessions_send behavior; otherwise agents may still behave as independent responders.