Chat SDK brings agents to where users already are

Chat SDK is a TypeScript library built to move agents into the chat platforms where users already work. It handles Slack, Microsoft Teams, Google Chat, Discord, Telegram, GitHub, Linear, and even WhatsApp from a single codebase, with the core chat package handling event routing and application logic.

The starting point was simple. As soon as internal teams built an agent, they had to implement separate integrations for platforms like Slack, Discord, GitHub, and Linear each time, and this quickly became repetitive work. What was needed wasn't bringing users to the agent, but a common layer that sends the agent to where users already are.

Chat SDK absorbs platform-specific differences through adapters.

• Slack renders formatting in real time via a native streaming path.
• Other platforms process streamed text through each adapter's markdown-to-native conversion pipeline.
• The Table() component renders to fit each platform: Slack's Block Kit table, Teams' and Discord's GFM markdown table, Google Chat's monospace text widget, Telegram's code block, and GitHub's and Linear's existing markdown pipelines.
• Cards, modals, and buttons only need to be written once in JSX, and they automatically fall back depending on each platform's level of support.

Message context is also automatically enriched. Channel names and usernames are converted into human-readable text, and conversely, when an agent sends an at-mention in clear text, it's converted back so that Slack notifications are actually triggered. Link previews, referenced posts, and images are also included in the prompt, and standard markdown is converted in real time into Slack's variant syntax.

The connection with the AI SDK is also key. The post() function can directly receive an AI SDK text stream, so streaming LLM responses can flow into chat platforms without separate wiring. In other words, it handles model responses, UI rendering, and platform-specific representation differences all at once.

State management uses a pluggable state adapter.

• Redis and ioredis have been supported since launch.
• PostgreSQL is now also offered as a production-ready option.
• It runs on pg and raw SQL, automatically creating the necessary tables on first connection.
• It supports TTL-based caching, distributed locking, and namespaced state.

Finally, with the addition of WhatsApp support, the write-once model now extends to an even larger messaging platform. It supports messages, reactions, auto-chunking, read receipts, downloading images/voice messages/stickers, and location sharing, but due to WhatsApp's 24-hour messaging window constraint, it does not support message history, editing, or deletion.

Chat SDK is open source and currently in public beta. Agents that teams build no longer need to stay confined to a single platform — they can be deployed into the chat environments where users actually are.