Your agent works.
Your frontend should too.

The production SvelteKit frontend for LangGraph agents. Ship a secure, customizable UI without a TypeScript backend — OIDC-ready and built to stay maintainable for years.

Explore streaming, tool calls, and auth — no local setup required. Read the docs

terminal
$ proto install
$ cp .env.example .env
$ moon :dev :oidc-mock
✓ Ready — frontend, backend, and OIDC mock running
Built with
Svelte LangGraph

The gap

Building the agent is the easy part.
The frontend shouldn't be hard.

Python and LangGraph make agentic AI tractable. But the frontend landscape leaves you choosing between tools that break under maintenance pressure, lock you into opinionated stacks, or demand a separate TypeScript backend just to render a chat window.

Upstream UI changes break your customizations without warning
Codebases too complex or generated to meaningfully audit
TypeScript backends you didn't ask for sitting between you and your Python agent
Forked repos that drift from upstream and lose security coverage
Integrations that work in the demo but crumble months into production

Why svelte-langgraph

Built for the long haul.
Not the demo.

React has plenty of agent UIs. Svelte deserves a first-class LangGraph frontend — fast, joyful to customize, and built for production ownership.

Connect Python agents directly

Connects directly to your LangGraph server. No TypeScript backend required. Full streaming for tool calls and complex agentic workflows.

Pass security review. Stay maintainable.

OIDC built in. Auditable code for security teams — with long-term security support as a design goal, not a patch-and-pray cadence.

Customize every layer — no lock-in

Every component layer is independently configurable. Built on bits-ui and shadcn-svelte — override anything, extend everything.

Swap models without rewriting

All LLM providers via the OpenAI-compatible protocol — Anthropic, OpenAI, local models, Azure, and more.

Built with Svelte, bits-ui, shadcn-svelte, Proto, and Moon — pinned toolchains across Node and Python.

Getting started

Clone it. Run it. Ship it.

Today: clone or fork the repo. Soon: install as a package — no fork required.

01

Install toolchain

Install Proto in the repo root (`proto install`). Pins Moon, pnpm, Node, and Python per .prototools — one toolchain for the whole monorepo.

02

Configure & run

`cp .env.example .env`, then `moon :dev :oidc-mock` — frontend, backend, and OIDC mock with hot reload. Always include :oidc-mock for local auth.

03

Deploy to your stack

Deploy with the included config — Docker, your OIDC provider, your LangGraph server. Pull upstream security patches without the rewrite treadmill.

Who it's for

Built for builders.

Python AI developers

Stop spending sprints on React glue. Focus on the agent — connect a frontend that speaks LangGraph natively without a TypeScript-heavy stack.

Boutique AI agencies

Ship branded agent UIs you can customize, maintain, and hand over to clients with confidence.

SaaS platform teams

Embed OIDC-ready agent chat your security team can audit — built for the long haul inside your existing product.

CX & support teams

Deploy LangGraph-powered chatbots without a custom frontend team — fork, brand, and ship support agents on a budget.

The landscape

Know what you're choosing.

Every tool in this space makes different trade-offs. Here's an honest look at where svelte-langgraph fits alongside other solid options.

Comparison of agent UI tools across security, LangGraph support, customization, and community features
svelte-langgraphLangflow Chainlit Open WebUI Custom React
Security review friendly
LangGraph / Python native
Code quality & auditability
Security-first (OIDC built in)
Full streaming (tools + workflows)
Per-component customization
No JS backend required
Active large community
Broad feature surface

Every project here represents real effort from people who care about the ecosystem. Choose svelte-langgraph for quality, security, and long-term maintainability — when a production-grade UI you can audit and support for years matters more than the broadest feature set out of the box.

Streamlit excels at data apps and demos — a different category than production agent chat UIs.

Roadmap

Where we are. Where we're going.

We're building in the open. Here's what's shipped, what's in progress, and what's next.

Core chat UI with streaming

Shipped

Full message streaming and tool-call visibility in the chat UI

OIDC authentication

Shipped

Secure login with any OpenID Connect provider

Dark mode & theming

Shipped

Complete design system with light/dark mode and Tailwind CSS tokens

Multi-language support

Shipped

i18n with Paraglide — English, Dutch, and Hindi included

Conversation history

In progress

Persistent threads backed by LangGraph checkpoints

File upload & attachments

In progress

Send images, PDFs, and documents to your agent

Plugin system

Planned

Drop-in extensions for custom tool renderers, message types, and layouts

Installable package (no fork required)

Planned

Use svelte-langgraph as a dependency — npm install and go

Embeddable web component

Planned

Drop the chat UI into any app — Svelte or not — as an in-app component or web component with auth

Admin dashboard

Planned

Monitor agents, view usage, and manage users from a built-in admin panel

Built in the open

Help shape what ships next.

MIT licensed on GitHub. Try the live demo, open an issue, send a PR, or tell us what you built — contributors steer the roadmap as much as the maintainers.

Your agent is the product.
The frontend is how users trust it.

Open source. MIT licensed. Security-first. Fork it, configure it, and ship something you'll still be proud of next year.

MIT License · Maintained by SynergyAI · Built with Svelte + LangGraph