21st Agents SDK is the fastest way to add an AI agent to your app. Define your agent in TypeScript, deploy in one command, and embed a production-ready chat UI with Built-in streaming, session management, usage billing, and observability — so you can focus on what makes your agent unique, not infrastructure. Backed by Y Combinator (W26).
Hey Product Hunt! Serafim & Sergey here, co-founders of 21st.dev (YC W26).
Last year we built 21st.dev — a platform that 1.4M+ developers now use to
build UI for their AI apps and agents. Along the way we built our own AI
developer tools and agentic applications. They were loved by tens of
thousands of developers and gained strong open-source support (14K+ GitHub stars).
Like many teams, we started by building our own agents from scratch.
Then Claude Code arrived. Its reasoning, context handling, and developer
workflows were far ahead of anything we had built ourselves. It became
obvious: if your product isn't powered by something like Claude Code,
your agent will always feel limited.
But bringing that level of capability into your product is not simple.
You still have to build the agent UI, sandbox execution infrastructure,
streaming systems, observability — an entire platform just to run the agent.
So we built 21st Agents SDK.
The agent execution layer for modern AI products:
→ Define in TypeScript — agent({ model: 'claude-sonnet-4-6', tools: {...} }). That's your entire agent definition.
→ One-command deploy — npx @21st-agents/cli deploy. Sandboxed cloud execution, managed for you.
→ Drop-in React UI — — production-ready chat with streaming, tool execution rendering, and theming.
→ Built-in observability — Session replay, tool call traces, error tracking.
→ Your code stays in your codebase. Full control. Infrastructure just works.
Stop rebuilding agent infrastructure. Start building better AI products.
What are you building? Drop a comment — we'd love to help you scope it out.
Docs: 21st.dev/agents/docs
Website: 21st.dev/agents
One thing stood out: it’s positioned as a component registry, but the product actually looks closer to AI product infrastructure (ui + agents + sdk)
Framing it that way might resonate more with teams building AI apps
Had a few shadcn installs turn into local forks fast. 21st pulling components, blocks, and hooks into one stack feels useful, and a really clear diff or version flow would make that npx shadcn path hold up once teams start customizing.
The developer experience here looks really well thought out. Being able to define an agent in TypeScript and deploy it quickly with built-in UI and infrastructure could remove a lot of friction for teams experimenting with agent features.
I like that things like session management and observability are already handled so developers can focus on the actual agent logic.
Curious if teams are mostly using this to build internal tools first or embedding agents directly into user-facing apps.
Congrats on the launch.
