← Blog

Langflow vs Flowise: One Is Maintained, One Is Archived

November 18, 2026

AI Agents & ToolsTool ComparisonsNo-Code & App Builders

Langflow vs Flowise used to be a close call between two open-source, drag-and-drop builders for LLM apps and agents. It isn't close anymore. Flowise's GitHub repository is archived and the project reached end of life on August 31, 2026. Langflow is still shipping releases.

So the honest short answer depends on where you're standing:

  • Picking a builder for a new project? Choose Langflow, or look at another maintained tool. Starting on Flowise today means starting on code nobody upstream will fix.
  • Already running Flowise? Your instance still works. The question is how long you want to own its patches, and that's most of this article.
  • Just curious how they differed? The comparison below covers runtime, licence, builders, MCP, hosting and security, checked against each project's own repository and docs.

Flowise is described as it last shipped. Nothing here is a benchmark.

Quick Comparison

LangflowFlowise
Status (November 2026)Active, version 1.13Archived, end of life since August 31, 2026
OwnerIBM, through its DataStax acquisitionWorkday, which acquired it in August 2025
LanguagePythonTypeScript on Node.js
LicenceMITApache 2.0, with a commercially licensed enterprise directory
InstallDesktop app, Docker, Python package, sourcenpm, Docker, source
BuildersOne flow canvas, with agents as componentsAssistant, Chatflow, Agentflow
MCPServer and clientClient and server nodes
Custom logicEdit any component's PythonCustom code nodes in JavaScript
Security fixesShipped in new releasesStopped; some published advisories list no patch
GitHub starsAbout 157,000About 55,600

Star counts are read from each GitHub repository as of November 2026.

Where Each Project Stands Today

This is the section that decides most cases, so it comes first.

Flowise: archived, with a clear timeline

The Flowise team published a sunset notice with three dates:

DateWhat happened
July 29, 2026Announcement and code freeze. No new features, no pull requests reviewed.
August 2026Repository moved to public archive. Issues and pull requests locked, npm packages and Docker images marked deprecated.
August 31, 2026End of life. The core team left Discord and GitHub.

The notice gives a reason in the team's own words: developers increasingly rely on coding agents for complex tasks, and a rigid low-code workflow "quickly hits the limit" on complexity.

The Flowise repository now carries an archive banner. The code stays visible, and the notice encourages teams to fork it.

About a year earlier, in August 2025, Workday announced it had acquired Flowise to bring an agent builder into its platform. The sunset notice doesn't mention Workday, so don't read a cause into the timing. Both facts are simply true.

One thing the notice leaves open is the hosted service. It talks about the open-source repository only, and the Flowise homepage still shows cloud plans next to the sunset banner. If you pay for Flowise Cloud, ask the vendor directly before you plan around it.

Langflow: active, under IBM

Langflow is on version 1.13. Its latest patch release on GitHub is dated November 17, 2026, and its docs already carry a preview of the next version.

On ownership: IBM announced in February 2025 that it would acquire DataStax, the company behind Langflow, and said it would keep supporting the open-source Langflow community. Today IBM lists the product on its own site as IBM Langflow. The code is still on GitHub under the same organisation and the same licence.

Corporate ownership cuts both ways. It brings paid support and a roadmap tied to IBM's watsonx products. It also means the project's direction is set by a large vendor. Flowise's history is the reminder that an acquisition guarantees nothing in either direction, so judge Langflow by its release cadence, which you can check any day on GitHub.

How They Differ as Builders

Python vs Node.js

This was always the cleanest dividing line.

Langflow is Python. Its docs call it an "open-source, Python-based, customizable framework." Every component is Python you can open and edit in the canvas. If your team already writes Python for data or ML work, custom components feel native.

Flowise is TypeScript on Node.js. You install it with npm and need Node 20 or later. Custom logic goes into JavaScript code nodes. That suited teams with a JavaScript web stack, and Flowise shipped an embedded chat widget for dropping a bot into a site.

Installation differs too. Langflow's install docs list four routes, and recommend Langflow Desktop, a standalone app for macOS and Windows that manages dependencies for you. Flowise's README lists npm and Docker, so it runs as a local server or a container.

One canvas vs three builders

Flowise split its interface into three builders, per its docs:

  • Assistant, the beginner route: instructions, tools and uploaded files for retrieval.
  • Chatflow, for single-agent systems, chatbots and simple LLM flows.
  • Agentflow, the superset, for multi-agent systems and workflow orchestration. The docs describe Agentflow V2 as an architecture built from standalone nodes, and list V1 as deprecating.

Langflow has one canvas. A flow is a graph of components, an agent is one of those components, and other flows or components can be attached to it as tools. There's also a Playground for testing a flow step by step.

Neither design is better in the abstract. Flowise's split was friendlier on day one. Langflow's single model has fewer concepts to learn once you're past day one.

MCP support

Both projects adopted the Model Context Protocol, in different shapes.

Langflow works in both directions. As a client, an agent in a flow can call tools from any MCP server. As a server, Langflow's MCP server docs describe exposing your flows as tools that MCP clients can use, over streamable HTTP with SSE as a fallback. That second direction matters: a flow you built can become a tool inside Claude, Cursor or another agent without extra code.

Flowise's docs list "MCP client/server nodes" among its integrations, so agents could reach MCP tools from inside a flow. That code is frozen at its last release. The protocol keeps moving, and Flowise won't follow it.

If MCP itself is new to you, Claude Skills vs MCP explains what it does and doesn't cover.

Licence and Hosting

Langflow's LICENSE is MIT, with no extra clauses.

Flowise's licence file is Apache 2.0 for everything outside the enterprise directory. That directory, and a few files with explicit notices, fall under a separate commercial licence. This matters more now than it did a year ago. If you fork Flowise, the Apache 2.0 part is yours to maintain and ship. The enterprise code, which includes features such as SSO, isn't covered by that grant, and the company that sold the commercial licence has wound the project down. Check your position before you build on them.

On hosting:

LangflowFlowise
Run it yourselfDesktop, Docker, Python package, Kubernetes guidesnpm or Docker, on deprecated packages and images
Vendor-hostedIBM's product page names Langflow Cloud and watsonx Orchestrate as publish targetsCloud plans still shown on the homepage; future not addressed in the sunset notice
Paid supportThrough IBMNone from the core team since end of life

Neither vendor's hosted pricing is quoted here. Read it on their pages on the day you decide.

Security: Patched vs Frozen

Both tools run code. That's the point of them, and it makes both a target.

Langflow says so plainly. Its security docs call the Langflow UI "an IDE and code execution platform" and state that it doesn't enforce isolation between users inside one Langflow process. Tenant isolation is your job, at the infrastructure level. The project publishes security advisories on GitHub, including critical ones, and fixes land in numbered releases. Official packages and Docker images now disable automatic login by default. A self-hosted Langflow needs regular upgrades, like any server.

Flowise also published advisories on GitHub, and patched many of them in its final releases. The difference is what happens next. Several entries on its advisories page dated September 10, 2026, after end of life, list version 3.1.4 and earlier as affected and show "Patched versions: None." Version 3.1.4 is the last release. With the repository archived, nobody upstream will close those.

That is the practical meaning of end of life. The problem isn't that Flowise is less secure by design. It's that a newly reported hole stays open unless you fix it in your own fork.

If You Already Run Flowise

You have three realistic paths.

1. Stay on a fork, on purpose. Reasonable for an internal tool that isn't reachable from the internet and rarely changes. Pin the version, read the open advisories against the features you use, back up your database and flow exports, and name an owner. Treat it as a bridge with an end date.

2. Move to Langflow. It's the closest match for Chatflow and Agentflow builds. Go in with the right expectation: this is a rebuild, not an import. Langflow's import and export docs cover Langflow's own JSON files and don't mention Flowise. Your prompts, tool choices and vector store carry over as decisions. The graph, any JavaScript code nodes and the embedded chat widget have to be redone. Budget for rewriting custom JavaScript in Python.

3. Move to something else. A document chatbot may fit a finished-app platform better, which Dify vs Langflow walks through. Flows that outgrew the canvas may belong in code. The full map by use case is in our Flowise alternatives guide.

Whatever you choose, inventory first. List each flow, who uses it, what it connects to, and whether it faces the internet. Internet-facing flows move first.

Which Should You Pick?

Your situationPick
New project, want a visual builderLangflow
Python team, custom componentsLangflow
JavaScript-only team, new projectLangflow with Python help, or a maintained Node-based tool
Running Flowise internally, low riskA pinned fork for now, with a migration date
Running Flowise on the public internetMigrate soon
Want flows callable from Claude or CursorLangflow's MCP server

And if Langflow itself doesn't fit, Langflow alternatives covers the other maintained options.

The Step Before Either

Flowise's shutdown note says builders are moving from low-code canvases to coding agents. There's a simpler version of the same thought: plenty of people who reached for a flow canvas only wanted the result.

If you'd rather run a tool someone already built than maintain a builder yourself, that's where Taku comes in. Taku Apps and Sites, built by the best AI minds. Find one, run it, and keep it in your Mac Dock. Taku is in Beta, and the Mac app is available now.

FAQ

Is Flowise shut down?

The open-source project is. The team froze the code on July 29, 2026, archived the GitHub repository in August, and reached end of life on August 31, 2026. The code remains public under Apache 2.0, so existing installs keep running and anyone can fork it. No upstream fixes are coming.

Is Langflow better than Flowise?

For a new project in November 2026, yes, for one reason that outweighs the feature list: Langflow is maintained and Flowise isn't. On features alone they were close. Langflow is Python-first with one canvas and a desktop app. Flowise was Node-based with three builders and an embedded chat widget.

Can I import Flowise flows into Langflow?

Not directly. Langflow imports and exports its own JSON flow format, and its docs don't describe a Flowise importer. Plan to rebuild each flow by hand, reusing your prompts, model settings and data sources.

Who owns Langflow and Flowise?

Langflow came from DataStax, which IBM announced it would acquire in February 2025. IBM now lists it as IBM Langflow. Flowise was acquired by Workday in August 2025.

Is Langflow open source?

Yes. The repository is MIT licensed. IBM offers support for it and names Langflow Cloud on its product page, but the builder itself can be self-hosted without a licence fee.

Is it safe to keep using Flowise?

It depends on exposure. An internal instance behind your network, on a pinned version, with an owner, is a manageable risk for a while. An instance reachable from the internet is harder to defend, because GitHub lists advisories against the final release with no patched version.

Does Langflow have a cloud version?

IBM's product page names Langflow Cloud and watsonx Orchestrate as places to publish flows. Langflow's own docs focus on Desktop and self-hosted deployment. Check IBM's page for current terms.