Building OpenClaw from Scratch: Integrating an AI Assistant into Feishu

OpenClaw Overview

OpenClaw is an open‑source AI‑agent runtime that equips a large language model with a "body": the ability to read/write files, manage messages, and automate routine tasks. The framework runs entirely on the user’s machine, keeping all data local, and can be extended with custom skills such as weather queries, document handling, and voice replies.

Local‑first execution : No data leaves the host computer.

Multi‑platform integration : Built‑in adapters for Feishu, WeChat, Telegram, Discord and other messaging services.

Model‑agnostic design : Any LLM can be plugged in; the author demonstrated the setup with Alibaba Tongyi Qianwen.

Memory system : Provides short‑term and long‑term stores so the assistant can remember user preferences across sessions.

Skill system : Developers add functions (e.g., weather lookup, document read/write, permission management) as independent modules.

Environment Preparation

System requirements

Operating system: Windows, macOS, or Linux.

Node.js 18+ (the author used v24.2.0).

Access to a large‑model endpoint – either a locally deployed model or a cloud API.

Common pitfalls and mitigations

Windows PowerShell may reject script execution. Resolve with Set-ExecutionPolicy RemoteSigned.

The OpenClaw gateway listens on port 3000 by default; if the gateway fails to start, verify that the port is free.

Feishu’s callback verification requires a correct response to the challenge parameter. OpenClaw handles the response automatically, but the host machine must be reachable from the internet.

The bot replies only when it is @‑mentioned or in a private chat; in group chats the bot must be added manually.

Integration Workflow with Feishu

Configure the gateway to expose an HTTP endpoint (default http://localhost:3000) and set the model‑access credentials in the environment configuration.

Register the gateway URL in the Feishu developer console as the bot’s callback address. Feishu will send a challenge request during verification; OpenClaw returns the required echo value.

Deploy the bot to a Feishu workspace, then add it to the desired chats. Ensure the bot is @‑mentioned for it to process messages.

Enable the built‑in skill modules:

Calendar reminder : The assistant periodically queries the user’s Feishu calendar and pushes upcoming events.

Email assistant : Unread emails are fetched via the configured mail API and summarized or organized automatically.

Weather notification : A weather‑lookup skill retrieves forecast data and sends a reminder before the user leaves.

Document management : The bot reads from and writes to Feishu cloud documents, enabling AI‑driven editing.

Voice reply : Text‑to‑speech integration generates audio messages that are sent back to the chat.

Monitor the gateway logs for any port conflicts or network‑reachability errors, and adjust firewall or NAT settings accordingly.

Outcome and Trade‑offs

Running the assistant locally eliminates the privacy concerns of cloud‑only bots and gives the user full control over data flow. The trade‑off is the additional setup complexity: configuring the gateway, handling network callbacks, and ensuring the correct execution policy on Windows. The author reports that once the bot successfully replies in Feishu, the effort is justified.