Feishu bot initialization: identity & memory

What this page covers

You can already talk to OpenClaw in Feishu, but it is not yet in a long-term collaboration state. Without initialization you often see unstable response style, poor understanding of your work habits, and claims it "remembered" without actually writing to the workspace. Initialization writes these long-term rules into the agent's workspace.

Before starting: verify three things

1. You can chat in Feishu — Deployment, bot creation or pairing, and a basic message loop are done.
2. The agent can read and write files — Initialization is a workspace configuration step, not a chat game.
3. Your goal is a baseline for long-term collaboration — If you plan to use it seriously, do it now.

Core files for first-time initialization

For the first run, create or update: AGENTS.md (identity and role), USER.md (who you are, language, work rhythm), SOUL.md (personality), and a greeting message.

Step 1: Confirm it can write files

Send in Feishu:

We are starting your formal initialization. Create or update AGENTS.md, USER.md, SOUL.md. Confirm you can read/write workspace files and read back content to me. If yes, reply READY and start with AGENTS.md.

Step 2: Set up identity with AGENTS.md

Write to AGENTS.md, then read and paste back to me:

# AGENTS
name: Lobster
role: Feishu work partner
type: self-hosted OpenClaw agent
vibe: Professional, direct, reliable

identity_description:
- You are an OpenClaw bot in Feishu
- You help with work, information, and methodology
- Your output should be clear, accurate, and actionable

Step 3: Configure long-term memory (USER.md)

Write to USER.md with your preferred_name, timezone, language_preference, working_profile, and collaboration_preferences. Adapt the values to your situation.

Step 4: Define personality with SOUL.md and set greeting

Use SOUL.md for tone, behavior, and boundaries. Configure a greeting the bot sends when a user starts a conversation (via Feishu app settings or AGENTS.md).

Step 5: Test the minimal conversation loop

Have the bot respond to a simple task. If it succeeds using workspace context, the loop works.

Final verification

Send: "Run an initialization check. List the core files, read and paste each file's full content, confirm they are written to workspace, and reply READY when done." It is complete only when the bot lists files, reads each back, content matches your intent, and it replies READY.

External links

• OpenClaw docs
• OpenClaw GitHub
• Feishu open platform
• Channels overview
• Feishu channel guide