@hwchase17 @LangChain I built an open-source CLI that scaffolds, runs, and deploys production LangChain agents โ frontend and backend in one command.
Every agent project starts the same way: wire up a LangGraph backend, scaffold a Next.js frontend, figure out CORS, hide API keys, set up a proxy. Two days gone before you write a single line of agent logic.
langctl kills that.
langctl new my-agent
cd my-agent
langctl dev
That's it. Agent + chat UI running at localhost:3000. One URL. No CORS. No exposed keys.
How it actually works:
langctl new scaffolds a LangGraph Agent Server backend and a Next.js frontend, wired together through a built-in proxy. The browser only ever talks to /api/agent/* on the frontend's origin โ the proxy attaches x-api-key server-side. Zero cross-origin requests, ever.
langctl dev starts both processes, waits for the agent's health endpoint before booting the UI, and tears everything down cleanly on Ctrl-C. Like next dev, but the backend is an agent.
Three chat UI options out of the box:
assistant-ui (default) โ threads, branching, composer runtime
minimal โ one hand-written Chat.tsx, zero UI deps
ai-elements โ shadcn-registry components (experimental)
Other commands:
langctl sync โ regenerates langgraph.json from agent.yaml, preserves hand-written keys
langctl doctor โ verifies toolchain, ports, keys, and config before something breaks
agent.yaml is your single source of truth. Dev and production differ by one env variable. The frontend source doesn't change.
Phase 1 (scaffold + unified dev runtime) is live on PyPI. Deploy providers (LangSmith Cloud, Vercel) are next.
Apache-2.0 licensed. Generated projects carry no license obligation to this tool.
GitHub:
PyPI:
pip install langctl and build your next agent in minutes, not days.
LangChain Langchain Community for Developers, India LangChain JS LangChain Developer LangChain Labs Formation LangChain FranceLangchain
#
OpenSource# #
LangChain# #
LangGraph# #
AIAgents# #
Python# #
DeveloperTools# #
CLI# #
FastAPI# #
NextJS#