// the find
vercel/chat
A unified TypeScript SDK for building chat bots across Slack, Microsoft Teams, Google Chat, Discord, and more.
A unified TypeScript SDK that abstracts away platform differences when building bots for Slack, Teams, Discord, Telegram, Google Chat, and others. Write your event handlers once against a common API; adapters translate to each platform's native wire format. Aimed at developers who need to ship the same bot to multiple chat platforms without maintaining separate codebases.
The adapter model is well-thought-out — separating platform adapters from state adapters (Redis, Postgres, memory) means you can swap persistence independently of the chat platform. JSX-based card authoring is a genuinely better experience than hand-authoring Slack Block Kit JSON or Adaptive Cards XML. The streaming support with platform-aware fallbacks (native Slack streaming vs. post+edit) handles the messiest part of AI bot UX correctly. The `llms.txt` / `llms-full.txt` and bundled agent skill are a practical addition — AI coding agents actually know the APIs before writing code.
This is a Vercel project building on a namespace they grabbed (`chat` on npm), which raises questions about long-term stewardship if Vercel pivots. Feature parity across platforms is the classic problem with abstraction layers — the feature matrix exists but a bot that relies on modals or slash commands will hit platform-specific gaps quickly. No mention of webhook signature verification handling in the README, which is table stakes for production bots and often done wrong. Community adapters (Mattermost, WhatsApp via Baileys) are third-party with no quality bar stated, so you're trusting unknown maintainers for critical integrations.