Upstreet – Open-source AI agent framework built with React

TL;DR

TL;DR: Upstreet is an open-source framework that lets you build AI agents using React components. Install the usdk npm package, scaffold a project with pnpm, and deploy agents that combine React UI patterns with LLM-powered reasoning.

Source and Accuracy Notes

⚠️ This section is MANDATORY. All links must be verified from actual source, not guessed.

• Project page: upstreet.ai ← MUST visit and verify
• SDK docs: docs.upstreet.ai ← verified via direct fetch
• Source repository: github.com/UpstreetAI/upstreet-core ← verified via direct fetch
• npm package: usdk on npmjs.com ← v0.0.110 confirmed
• License: GPL-3.0-only (verified from packages/usdk/package.json license field)
• HN launch thread: news.ycombinator.com/item?id=42230070 (3 points, Show HN)

What Is Upstreet?

Upstreet is an open-source framework for building AI agents with React. Rather than writing agents as plain Python or TypeScript classes, Upstreet treats agents as React components — complete with props, state, rendering, and composition. The framework is designed around the idea that agent behaviors can be expressed as reactive UI-like structures, with tools and memory wired in as first-class primitives.

The core SDK is published as usdk on npm. According to the README, the framework is built around a monorepo containing:

• /apps/chat — a reference chat application
• /apps/docs — documentation site
• /packages/usdk — the core TypeScript SDK

The README describes Upstreet as:

“an open source framework to build + deploy AI agents with React.”

The SDK is in public beta. The docs banner confirms: “The Upstreet Agents SDK is now in public beta.”

Setup Workflow

Step 1: Install Node.js and pnpm

Upstreet requires Node.js 22.9.0 or above and uses pnpm as its package manager.

# Check your Node version
node --version

# Install pnpm if you do not have it
npm install -g pnpm

# Verify pnpm
pnpm --version

Step 2: Install the USDK globally

pnpm i -g usdk

Verify the installation:

usdk --version
# Should print: 0.0.110 (or newer)

Step 3: Scaffold a new agent project

usdk create my-agent
cd my-agent
pnpm install

The usdk create command scaffolds a project with the agent structure, default tools, and a chat interface ready to run.

Step 4: Configure your LLM provider

Create a .env file in the project root:

OPENAI_API_KEY=sk-...
# or Anthropic
ANTHROPIC_API_KEY=sk-ant-...

The SDK supports multiple LLM providers. Consult the docs on provider configuration for the full list of supported backends.

Step 5: Run your agent

usdk dev

This starts a local development server with hot reload. Open the URL printed in the terminal to interact with your agent through the built-in chat UI.

Step 6: Build and deploy

When your agent is ready, build it for production:

usdk build

The output is a deployable bundle. You can self-host it or use a platform like Railway, Vercel, or Fly.io.

Deeper Analysis

Architecture

Upstreet’s monorepo structure gives you three entry points:

| Directory | Purpose | |---|---| | packages/usdk | Core SDK — agent primitives, tool registry, LLM adapters | | apps/chat | Reference UI — a complete chat interface built with the SDK | | apps/docs | Documentation site |

The SDK itself is a TypeScript workspace package. The agent model treats each agent as a stateful component that can render outputs, maintain memory across turns, and call tools — all expressed in TypeScript/React rather than a domain-specific language. Agents are structured as React components, with tools registered as composable units.

SDK versioning and stability

The npm package usdk is at v0.0.110, and the docs carry a public beta banner. This means the API surface is still subject to change. If you are building production agents, pin to a specific version in your package.json:

{
  "dependencies": {
    "usdk": "0.0.110"
  }
}

Contributing

The repository accepts contributions. To set up the development environment:

git clone https://github.com/UpstreetAI/upstreet-core.git
cd upstreet-core
pnpm install

Consult the README contributing section for the full dev setup.

Practical Evaluation Checklist

• [ ] Node.js 22.9.0 or above installed
• [ ] pnpm installed (npm install -g pnpm)
• [ ] usdk installed globally (pnpm i -g usdk)
• [ ] usdk --version returns a version number
• [ ] usdk create scaffolds a project without errors
• [ ] .env file configured with an API key
• [ ] usdk dev starts the local server
• [ ] Agent responds to a basic prompt
• [ ] usdk build produces a production bundle

Security Notes

• API keys — never commit your .env file. Add it to .gitignore before running usdk create.
• Self-hosting — Upstreet agents run entirely on your own infrastructure. No data leaves your servers unless you explicitly call an external LLM API.
• Dependencies — the SDK pulls in npm packages. As with any Node project, audit your dependency tree regularly with pnpm audit.

FAQ

Q: How does Upstreet differ from writing agents with LangChain or LlamaIndex? A: LangChain and LlamaIndex are primarily library-style abstractions over LLM calls and retrieval pipelines. Upstreet’s distinctive bet is treating agents as React components — meaning you compose agent behaviors with the same component model you use for building user interfaces. This is a different mental model rather than a direct competitor.

Q: Does Upstreet require a specific LLM provider? A: No. The SDK is provider-agnostic and configurable. Set your API key via environment variables. See the install docs for the current list of supported backends.

Q: Is Upstreet production-ready? A: The SDK is in public beta (v0.0.110). The API may change. For production use, pin to a specific version and test thoroughly after each upgrade.

Q: Can I self-host an Upstreet agent? A: Yes. The agent is a Node.js application you can deploy anywhere that runs Node — your own VPS, Railway, Fly.io, Vercel, or a container. The SDK itself is open-source under GPL-3.0-only.

Q: What is the license for the Upstreet SDK? A: The usdk package is GPL-3.0-only, verified from the package.json license field in the source repository.

Conclusion

Upstreet takes a distinctive approach to agent development: expressing agent behavior through React components rather than procedural scripts. The usdk package gives you a structured SDK with tool registration, memory primitives, and LLM adapter support. The public beta status means the framework is functional but still moving. If you are building agents and want a component-based mental model that mirrors how you build React UIs, Upstreet is worth an afternoon’s exploration.

Get started: docs.upstreet.ai/install — pnpm i -g usdk to install, then usdk create to scaffold your first agent project.