Build, run, and iterate on AI agents with no code using Langflow on your own machine. Langflow stands entirely on its own: you do not need a workflow engine, a cloud platform, or any integration to design real, working agents here.
If you want to learn a no-code AI agent builder that is independent of any workflow engine, this is the track for you. Connect an agent to a workflow engine later — through Cloudflare Tunnel and a Camunda 8 REST call — only if and when you want to embed it in a business process.
This workshop spans two repositories — each works on its own:
Repository What it teaches Langflow — no-code AI agent builder — you are here Build no-code AI agents, independent of any workflow engine. Camunda workflow — business loan onboarding Orchestrate an agentic business process on Camunda 8, and call your Langflow agent from it.
Use synthetic test data only. Keep all API keys out of Git, screenshots, and shared documents.
This repository is part of the code-along webinar “Build a No Code AI Banking Agent.”
| Session | Build a No Code AI Banking Agent (free to join) |
| When | Monday, June 29, 11 AM ET |
| Presenter | Anjali Jain — Enterprise AI Architect at Metro Bank, CTO at Erdos Research, and Senior Tutor in AI & Machine Learning at the University of Oxford. Author of AI-Assisted Programming for Web and Machine Learning and the forthcoming Enterprise Architecture in an Agentic World. |
The live session is 45 minutes of content followed by 15 minutes of questions:
| Segment | Focus |
|---|---|
| Session overview | The complete picture: what we build and why it matters in regulated environments |
| Langflow | Build a no-code AI agent — the core agent-building skill, standalone |
| Camunda | Orchestrate that agent inside an auditable business process |
| Q&A | 15 minutes |
You can follow the Langflow portion entirely on its own, even if you never touch a workflow engine.
- About this workshop
- Why Langflow
- What you will build
- Repository structure
- Prerequisites
- Getting started
- What the guide covers
- Environment variables
- Connecting to a workflow engine (optional)
- Security and safe data
- Reference documentation
- License
Langflow is a no-code / low-code visual builder for AI agents and flows. With it you can:
- Assemble agents, prompts, tools, and retrieval visually — no application code required.
- Connect any supported model provider: OpenAI in the cloud, or a fully local Ollama model for zero-cost, private experiments.
- Test agents interactively in the Playground and expose them through a clean API.
- Own the whole loop on your own laptop — build, run, and refine without depending on any external platform.
This is a complete skill in its own right. Everything below works whether or not you ever use Camunda.
Model provider
OpenAI API · or local Ollama
│
▼
┌─────────────────────────┐
│ Langflow (your laptop) │ ← build & run agents here, on their own
│ http://localhost:7860 │
└─────────────────────────┘
│
▼ (optional) expose via Cloudflare Tunnel
Connect to a workflow engine such as Camunda 8
The Langflow box is the hero. The tunnel and workflow-engine step are optional extensions for when you want an agent to participate in a business process.
langflow-local-setup-workshop/
├── README.md
├── LICENSE
├── .env.example # Reference environment variables (copy to .env)
├── docs/
│ └── local-setup-guide.md # Full setup: Desktop, venv, Ollama, Playground, API, tunnel
└── flows/ # Export your Langflow flow JSON here
└── README.md
| Item | Minimum |
|---|---|
| Operating system | Windows 10+ or macOS 13+ |
| Python | 3.10–3.14 (python.org/downloads) — only for the Python install route |
uv package manager |
Install guide — only for the Python install route |
| Model access | An OpenAI API key, or Ollama for local models |
| Camunda 8 account | Only if you later connect the agent to a workflow engine (console.camunda.io) |
The simplest route is Langflow Desktop, which manages Python and dependencies for you.
git clone https://github.com/CodeDaim0n/langflow-local-setup-workshop.git
cd langflow-local-setup-workshop- Follow the local setup guide.
- Confirm Langflow opens at
http://localhost:7860. - Build an agent and test it in the Playground.
- Iterate — add tools, prompts, and a model of your choice.
- Optionally export your flow JSON to
flows/(without any keys).
That is a complete, standalone agent-building loop. The next section is only for connecting it to a workflow engine.
| Topic | Section in the guide |
|---|---|
| Langflow Desktop (simplest route) | Route A |
Python uv virtual environment |
Route B |
| OpenAI API key | Model provider setup |
| Ollama + Qwen (local models) | Zero-cost local models |
| Build and test an agent | Playground |
| Expose an agent over HTTP | Langflow API key |
| Cloudflare Tunnel | Quick tunnel and named tunnel |
| Camunda REST Connector | Calling /api/v1/run/FLOW_ID |
| Troubleshooting | Connector and tunnel issues |
Copy .env.example to .env and fill in your own values. The .env file is git-ignored and must never be committed.
cp .env.example .envWhen you want an agent to run as a step in a business process:
- Create a Langflow API key and call your flow with the
x-api-keyheader. - Expose Langflow with a Cloudflare Tunnel to get a public HTTPS address.
- Store the key as a Camunda Connector Secret (
LANGFLOW_API_KEY) and call the flow from a REST connector.
The full Camunda business-process example lives in the companion repository: business-loan-onboarding-workshop.
- Never commit
.env, OpenAI keys, or Langflow API keys. - Do not embed API keys inside exported flow JSON.
- Use HTTPS through Cloudflare for any external call into Langflow.
- Stop
cloudflaredwhen you finish (Ctrl + C). - A local tunnel is for development only — never expose it as a production service.
- Langflow documentation
- Langflow Desktop
- Langflow API keys and authentication
- Ollama and model library
- Cloudflare Tunnel
- Camunda REST connector
If you like this repo, give it a star :) — it helps others find it.
If you find issues, please help others by providing corrections :) — open an issue or a pull request.
See LICENSE. These materials are provided for educational use as part of the workshop.