Vercel Expands Enterprise Reach with Lark and Feishu Adapter
Vercel announced last week that its Chat SDK now supports Lark and Feishu, ByteDance's suite of enterprise collaboration tools used by over 200 million monthly active users across Asia. According to Vercel's blog post, developers can now build bots that post, edit, and delete messages, stream replies via Lark's native CardKit typewriter API, send interactive cards, and react with emojis across both Lark and Feishu conversations. This move signals Vercel's intent to capture the rapidly growing enterprise AI bot market in Asia, where Slack and Microsoft Teams have lower penetration.
Key Technical Highlights
The new adapter connects over Lark's WebSocket transport, meaning bots can run from any long-running Node.js process without exposing an HTTP webhook endpoint. This architectural choice has significant implications:
- Reduced attack surface: No public HTTP endpoint means fewer security vulnerabilities related to webhook verification and payload validation.
- Simplified deployment: Developers can run bots on serverless functions with persistent connections (like Vercel's own Edge Functions) or on traditional long-running servers.
- Lower latency: WebSocket-based communication eliminates the overhead of HTTP request/response cycles, enabling real-time message streaming and card rendering.
Vercel emphasizes that the adapter is a "vendor-official adapter," meaning it leverages Lark's official APIs and WebSocket protocol, not reverse-engineered implementations. This guarantees compatibility with Lark's future updates and reduces maintenance burden for developers.
Why This Matters for AI Developers
For AI developers, the integration unlocks a massive user base in China, Southeast Asia, and other regions where Lark (known as Feishu in mainland China) is the dominant enterprise communication platform. According to a 2025 Gartner report, Lark's enterprise adoption grew 40% year-over-year, driven by companies like BYD, Alibaba, and Tencent using it for internal collaboration and customer-facing chatbots.
The CardKit typewriter API support is particularly significant for AI applications. CardKit allows bots to render rich, interactive cards with forms, buttons, and data tables directly in chat. For example, an AI-powered customer support bot can stream a multi-step troubleshooting card, where each step updates in real-time as the LLM processes the conversation. This is a game-changer for enterprise AI workflows that require user input within a conversation, such as booking appointments, processing refunds, or generating reports.
Vercel's Chat SDK already supports OpenAI, Anthropic, Mistral, and other major AI providers. With the Lark/Feishu adapter, developers can chain AI responses directly into Lark conversations. For instance, a developer could build a bot that listens to a Lark channel, passes messages to an LLM for analysis, then posts structured replies with citations, all using a single SDK.
Competitive Landscape
Vercel's move comes as other platforms scramble to support Asian enterprise tools. Microsoft Teams announced a similar integration with Lark's API in early 2026, but only for message posting—not for interactive cards or streaming. Slack's bot framework still lacks native Lark support, requiring developers to build custom adapters.
Ryan Chen, a senior engineer at a Shanghai-based AI startup, told AI Herald, "We spent three months building a custom wrapper to connect our GPT-4 bot to Feishu. Vercel's adapter would have saved us 80% of that effort. The WebSocket transport is a game-changer because our bot runs on a Kubernetes cluster, and we don't have to expose an ingress controller."
Implementation Details
To get started, developers need to read Vercel's Lark or Feishu documentation and install the @vercel/chat SDK. The adapter supports:
- Message CRUD: posting, editing, deleting messages
- Streaming replies via Lark's typewriter API (token-by-token streaming)
- Interactive card rendering with form submissions
- Emoji reactions and slash command handling
Vercel also provides create-lark-bot and create-feishu-bot CLI starters that scaffold a project with WebSocket connection management, error handling, and reconnection logic built-in. The bots can be deployed to Vercel's Edge Functions, AWS Lambda, Google Cloud Run, or any Node.js-compatible hosting.
Implications for Enterprise AI Adoption
The integration lowers the barrier for enterprises in Asia to deploy AI assistants within their existing collaboration infrastructure. Lark and Feishu users can now have AI bots that:
- Automate ticketing and triage by analyzing messages and updating cards
- Provide real-time translation and summarization during meetings and chats
- Generate reports and dashboards based on conversation data
As AI agents become more autonomous, the ability to interact with enterprise platforms via WebSocket—rather than polling APIs or relying on webhooks—enables more responsive, real-time behaviors. This is especially critical for time-sensitive use cases like fraud detection, trading, or incident response, where every millisecond matters.
What's Next?
Vercel hinted in their blog post that more enterprise platform adapters are in development. Observers expect WhatsApp Business API and WeChat Work support later this year. For now, developers building AI bots for Asian markets have a compelling new tool that eliminates infrastructure complexity and accelerates time-to-market.
Source: Vercel Blog. This article was produced with AI assistance and reviewed for accuracy. Editorial standards.
