Feishu Multi-Bot

Configure multiple Feishu/Lark bot identities within a single OpenClaw instance — each Agent gets its own Feishu App, name, avatar, and message routing. Use when: (1) Setting up multi-bot Feishu integration where each Agent appears as a separate bot, (2) Configuring message routing between Feishu apps and OpenClaw agents, (3) Debugging Feishu bot routing issues (bot not responding, wrong agent receiving messages, gateway startup failures), (4) Designing a multi-Agent architecture with per-role Feishu entry points, (5) Setting up per-project Feishu group isolation. Also trigger for: '飞书多机器人', '一个龙虾多个飞书', '飞书路由', 'Feishu accounts', 'Lark bot routing', or any question about making multiple Agents appear as separate bots in Feishu/Lark. Based on production experience running 10+ Agents with independent Feishu bots (2026-04).

OpenClaw Feishu Multi-Bot Configuration

One OpenClaw instance, multiple Feishu bot identities. Each Agent appears as an independent bot in Feishu — own name, own avatar, own group memberships. Users see separate assistants; behind the scenes, a single Gateway dispatches everything.

Architecture in 30 Seconds

OpenClaw Gateway (single instance)
  ├── Agent: orchestrator ←→ Feishu App 1 (总调度 bot)
  ├── Agent: content-writer ←→ Feishu App 2 (写作 bot)
  ├── Agent: code-expert ←→ Feishu App 3 (开发 bot)
  └── Agent: analyst ←→ Feishu App 4 (分析 bot)

Three configuration blocks must align:

Block	Location in openclaw.json	Purpose
channels.feishu.accounts	Feishu credentials per bot	Maps accountId → appId/appSecret
bindings	Message routing rules	Maps accountId → agentId
agents.list	Agent definitions	Maps agentId → workspace/model/tools

The accountId is the key that ties them together. Get it wrong in any one place and routing breaks silently.

Quick Start

1. Create Feishu Apps

One enterprise app per Agent at open.feishu.cn/app. Enable "Bot" capability. Record each app's AppID and AppSecret.

2. Run the setup helper

scripts/setup-feishu-bots.sh orchestrator:cli_xxx:secret1 writer:cli_yyy:secret2 coder:cli_zzz:secret3

Generates the three JSON blocks ready to paste into openclaw.json.

3. Register agents and restart

Add the generated config to openclaw.json, then:

openclaw doctor && openclaw gateway restart

4. Test each bot

Send a message to each Feishu bot independently. Verify each responds with the correct Agent identity.

Reference Files

File	Read when...
references/architecture.md	Understanding the three-block config model, accountId mechanism, and channel layer design
references/build-guide.md	Setting up from scratch — Feishu developer console through gateway restart, step by step
references/routing-deep-dive.md	Debugging routing issues — accountId consistency checks, binding rules, group-based isolation
references/troubleshooting.md	Fixing specific problems — gateway won't start, bot not responding, wrong agent, spawn conflicts

Critical Lessons (Save Hours)

1. accountId must match in 3 places: channels.feishu.accounts.{key}, bindings[].match.accountId, and the agent field inside the account config. One typo = silent routing failure.

2. Binding type must be "route": Using "delivery" or any other value causes gateway startup failure with no helpful error message.

3. Feishu apps must be published: Draft-state apps cannot receive messages. This is the #1 "bot not responding" cause.

4. allowAgents must be complete: If your orchestrator spawns sub-agents, every spawnable agent ID must be in allowAgents. New agent added but not listed = spawn permission error.

5. agentToAgent must stay OFF: Enabling agentToAgent.enabled: true breaks all sub-agent spawning (known bug #5813). Keep it false.

See references/troubleshooting.md for the full diagnostic flowchart.

Scripts

Script	Usage
scripts/setup-feishu-bots.sh	./setup-feishu-bots.sh <agentId:appId:appSecret> ... — Generates channels, bindings, and agents.list JSON blocks