Langflow is a free, open-source, Python-based tool for building AI agents and workflows by dragging and connecting components on a visual canvas. You can build chatbots, RAG apps (chat with your own documents) and multi-step agents, then run each flow as an API or as an MCP tool. It is low-code, not no-code: you can start visually, but every component is editable Python. Langflow is now owned by IBM through its DataStax acquisition, and it stays MIT-licensed.
Quick answer
- What it is: a visual builder for AI agents and RAG pipelines, built with Python.
- Cost: free to self-host under the MIT license.
- Install:
uv pip install langflow, thenuv run langflow run. Opens at http://127.0.0.1:7860. - Needs: Python 3.10 to 3.14 and uv. Minimum 2 GB RAM and a dual-core CPU; 4 GB RAM or more is recommended.
- Works with: major LLMs, vector databases, APIs, and MCP (Model Context Protocol) as both client and server.
- Who owns it: IBM, after its acquisition of DataStax, which had bought Langflow in 2024.
- Not the same as: Zapier or Make. Langflow is for building AI agents and apps, not for general business automation.
What is Langflow?
Langflow is an open-source platform for building and deploying AI-powered agents and workflows. You build on a canvas: each block (called a component) does one job, such as taking chat input, calling a language model, searching a vector database or running a tool. You connect the blocks with lines to show how data moves. The result is called a flow.
Two features separate it from simple drag-and-drop tools:
- Code access: you can open any component and change its Python code.
- Built-in API: every flow can run as an API endpoint, so your website or app can call it, and it can also act as an MCP server so other AI tools can use it.
A note on older descriptions: earlier guides, including our first version of this article, described Langflow as a no-code automation tool for business tasks. That is not accurate. It is a developer-friendly, low-code builder for AI agents and RAG apps.
What can you build with Langflow?
| Use case | What the flow does |
|---|---|
| Chatbot | Takes a question, sends it to an LLM with a prompt, returns an answer |
| RAG app (chat with your documents) | Loads files, splits and embeds them, stores them in a vector database and retrieves the right pieces to answer questions |
| AI agent with tools | Lets the model decide when to search the web, call an API or query data |
| Multi-agent workflow | Several agents with different roles pass work to each other |
| MCP tool | Exposes a flow as a tool that other MCP-compatible apps can call |
| Prototype for clients | Build visually, test in the built-in Playground, then hand over an API |
Key features of Langflow
- Visual builder. Drag, drop and connect components, then test the result immediately.
- Component library. Hundreds of ready components for models, prompts, data loaders, vector stores, agents and tools.
- Editable Python. Open any component and customise it.
- Model and database neutral. Use OpenAI, Anthropic, Google, IBM Granite, local models through Ollama, and many vector databases.
- Agents and tools. Add an Agent component and give it tools it can choose from.
- MCP support. Connect to external MCP servers for tools, and expose your own flows as MCP tools.
- API and export. Call a flow over REST, or export it as JSON for use in a Python app.
- Playground. A built-in chat window to test the flow as you build.
- Triggers and admin features. Recent releases have added scheduled and webhook triggers and team administration, according to release coverage; check the release notes for your version.
How to install Langflow
The recommended way is uv, a fast Python package manager.
uv venv langflow-env
source langflow-env/bin/activate
uv pip install langflow -U
uv run langflow runOn Windows, activate the environment with langflow-env\Scripts\activate.
Open http://127.0.0.1:7860 in your browser. The official release packages turn automatic login off, so set a password before you start:
LANGFLOW_SUPERUSER_PASSWORD=choose-a-strong-passwordPut it in a .env file and start with uv run langflow run --env-file .env, then sign in with the username langflow.
Other ways to run it:
- Docker: run the official image, with Postgres if you want a production-style setup.
- Langflow Desktop: a downloadable app for easy local use. The documentation notes that the desktop version lacks some features, such as the shareable Playground and voice mode.
- No official hosted cloud: DataStax’s hosted Langflow was reported shut down in April 2026, so plan to self-host.
If you are new to Python, practise first with our Python tutorial and the online Python compiler.
How to build your first flow in Langflow
- Start Langflow and open it in your browser.
- Create a new flow and pick a starter template (for example a simple agent or a chatbot), or start from a blank canvas.
- Add a language model. Choose a provider and paste your API key into the component, or connect a local model through Ollama.
- Write a prompt. Add a Prompt component and tell the model its role and rules.
- Connect input and output. Link the Chat Input to the model and the model to the Chat Output.
- Add tools or documents. For an agent, add tools such as web search or a calculator. For RAG, add a File component, a text splitter, an embedding model and a vector store.
- Test in the Playground. Ask real questions and watch which steps run.
- Fix and improve. Adjust prompts, chunk sizes and tools until answers are reliable.
- Deploy. Call the flow through its API, or expose it as an MCP tool.
Write good prompts for each step; our guide to prompt engineering in AI will help.
Langflow vs other tools
| Tool | Type | Best for | Note |
|---|---|---|---|
| Langflow | Visual builder for AI agents and RAG, Python | Developers who want visual speed plus code control | Free, MIT license, owned by IBM |
| Flowise | Visual builder, Node.js | JavaScript teams | Reported archived in August 2026 after its owner announced a sunset; confirm before choosing |
| Dify | AI app platform | Teams wanting a full app platform | License has extra conditions; read it before commercial use |
| n8n | Workflow automation | Connecting apps and business processes, with AI steps | General automation, not AI-first |
| Zapier and Make | No-code automation | Business users linking apps | Not built for custom AI agents |
| LangChain | Code framework | Developers writing everything in code | Langflow began as a visual layer for it |
Use Langflow when you want to prototype AI agents quickly and keep full control of the code. Use Zapier or Make when you want to connect business apps without any coding.
Is Langflow free?
Yes. The open-source software is free under the MIT license, and you can self-host it with no limit on flows or users. You will still pay for the services your flows call, such as LLM API usage and any vector database. For enterprises, IBM offers paid support options, and there is no public price list.
Is Langflow secure?
Treat it like any developer tool that can run code:
- Keep it updated. Langflow has had serious security issues, including a critical remote code execution flaw in 2025 that was fixed in version 1.3.0 and later exploited in attacks.
- Do not expose it to the internet without authentication. Turn off automatic login and set a strong superuser password.
- Protect API keys. Store them as environment variables or secrets, never inside shared flows.
- Review community components before using them, since components are Python code.
- Back up before upgrading, as the docs recommend.
For more on running software safely on your own server, read our guide on how to check if your PHP website was hacked.
Common mistakes to avoid
- Treating Langflow as a no-code tool and skipping Python basics
- Putting an API key directly in a flow you plan to share
- Starting with a big multi-agent design instead of one simple flow
- Not testing with real questions and edge cases
- Forgetting to set a password before running it on a server
- Ignoring the costs of LLM calls, which grow with usage
Related reading
To learn about related AI tools, see our guides on AI tools to chat with large PDF files, top AI frameworks for building apps, what are large language models and free AI tools for students in India.
FAQs
What is Langflow used for? It is used to build AI chatbots, RAG apps, AI agents and multi-agent workflows with a visual editor, then run them as APIs or MCP tools.
Is Langflow free and open source? Yes. It is open source under the MIT license, and you can self-host it for free.
Do I need to know coding to use Langflow? You can build simple flows visually, but basic Python knowledge helps a lot, because every component is Python code and real projects need customising.
What Python version does Langflow need? Python 3.10 to 3.14, with uv recommended for installation.
What is the difference between Langflow and LangChain? LangChain is a code framework. Langflow is a visual builder that lets you assemble similar pipelines on a canvas and still edit the Python behind them.
Who owns Langflow? IBM, through its acquisition of DataStax, which had acquired Langflow in 2024. The project remains open source.
Is Langflow better than n8n? They solve different problems. Langflow is for building AI agents and RAG apps. n8n is for automating workflows across apps, with AI as one part.
Can Langflow work with local models? Yes. You can connect local models, for example through Ollama.
Does Langflow support MCP? Yes. It can connect to external MCP servers as a client and expose your flows as an MCP server.
Is there a hosted version of Langflow? Reports say DataStax’s hosted Langflow cloud was shut down in April 2026. The recommended route is to self-host the open-source version.
Conclusion
Langflow is a practical way to turn an AI idea into a working agent or RAG app quickly: build visually, keep the Python under your control, and ship as an API or MCP tool. Learn basic Python first, start with one small flow, secure your installation, and keep it updated.
Details were checked on October 29, 2026. Langflow releases often, so confirm the latest version, Python requirements and features in the official documentation.