OpenClaw launched its second messaging channel this week: Feishu. This move is noteworthy—open-source AI Agents (AI programs capable of autonomously executing multi-step tasks) are no longer confined to geek circles like Telegram. Instead, they are embedding directly into the work windows where Chinese white-collar workers spend 8 hours a day online.
What this is
OpenClaw is an open-source AI Agent framework codenamed "Crayfish." Its core logic is multi-channel access: build the AI capabilities first, then connect them to various chat tools. It previously only supported Telegram; this week, it added Feishu.
There are two key details to the Feishu integration. First is the security model: private chats from strangers don't go directly into the LLM for processing. Instead, a pairing code is generated, requiring manual approval by an administrator in the terminal—this is essential for enterprise scenarios. Second is the technical approach: it uses a WebSocket persistent connection (the local machine actively connects to the Feishu server, rather than exposing a port and waiting for Feishu to connect). You don't need to buy a public network server; running it locally allows you to receive messages, making it friendly for small and medium-sized teams.
Post-integration use cases: @-ing the AI in a group to do research, privately messaging it to organize daily reports, or directly reading and writing Feishu docs to extract meeting minutes. The core logic is that message traffic is already in Feishu; AI must follow the flow there so it doesn't interrupt the work rhythm.
Industry view
We note an emerging consensus: the deployment bottleneck for AI Agents lies not in model capability, but in "where it lives." An independent webpage or command line means you must proactively seek out the AI, but embedding it in IM means the AI is right where you work. Dust and Q in the Slack ecosystem have already validated this path; Feishu is the most likely comparable scenario in the Chinese market.
However, the opposing voices are equally clear. First is data compliance: routing enterprise chat logs and documents to third-party AI for processing is a red line for many companies. Although OpenClaw is open-source and locally deployable, a security mechanism like "pairing code approval" remains too crude for enterprise-grade requirements. Second, Feishu is pushing its own AI capabilities. How far third-party Agents can go in the Feishu ecosystem is an open question—domestic developers are no strangers to platforms acting as both referee and player. Finally, the enterprise approval process for self-built apps is typically slow, which inherently acts as a hidden threshold for adoption.
Impact on regular people
For enterprise IT: There is now an additional open-source, locally deployable AI assistant option, but they must independently assess the risks of data flowing through third-party models. Feishu document read/write permissions, in particular, must be granted cautiously.
For individual professionals: The trend of AI shifting from a "standalone tool" to a "group chat member" is clear. Having an AI collaborator in a project group will become the norm—learning to @ it to assign tasks is a new collaboration skill currently taking shape.
For the consumer market: The short-term impact is limited. OpenClaw's target users remain technical teams and early AI adopters; it is still a considerable distance away from ordinary consumers.