Analytics cookies only (Google Analytics, PostHog). No ad networks, no cross-site tracking. See our privacy policy.
upstream: github.com/FlowiseAI/Flowise (opens in new tab) · github.com/langflow-ai/langflow (opens in new tab)
Pick Flowise
If you already have flows on it or prefer its TypeScript stack and lighter idle RAM (~697 MB against ~1.2 GB), and are willing to run a frozen release or maintain a fork of its Apache-2.0 code.
Pick Langflow
If you want a repository that has not been archived, Python components you can edit and a plain MIT licence.
Flowise and Langflow are two visual builders for LLM applications: drag components onto a canvas, connect a model, a vector store and some tools, then call the result through an API. Langflow is Python and MIT-licensed; Flowise is TypeScript and Apache-2.0 with commercially licensed enterprise modules. There is a project-status difference you should know before you pick: Flowise's repository has been archived by its owner. If you are also weighing Dify, see Dify vs Langflow.
Flowise's docs call it "an open source generative AI development platform for building AI Agents and LLM workflows". A Chatflow "is designed to build single-agent systems, chatbots and simple LLM flows"; an Agentflow can build "multi-agent systems, and complex workflow orchestration"; an Assistant is the beginner-friendly option that answers from uploaded files. It offers API access, JS and Python SDKs and an embeddable chat widget.
Langflow describes itself as "a powerful platform for building and deploying AI-powered agents and workflows". Its docs call it "an open-source, Python-based, customizable framework", where you can "Create your own Python components", "Expose Langflow as an MCP server", connect to external MCP servers, and "Use the Langflow API to run flows from your application code".
The Flowise README now opens with "Flowise has been archived", and the GitHub banner says the repository is read-only. The maintainers' announcement says they "decided to wind down our operations for Flowise", that feature development stopped at a code freeze, that issues and pull requests would be locked on archival, and that npm packages and Docker images would be marked deprecated. It also says "the Apache 2.0 licensed code is yours to keep building on" and encourages forks.
For a self-hoster, that means the announcement says new pull requests are no longer reviewed. The install we verified still runs, but treat it as a frozen snapshot unless you take on a fork.
Flowise's README says its source code is "made available under the Apache License Version 2.0". Our catalog lists it as Apache-2.0 core with commercially licensed enterprise modules. Langflow's README shows an MIT licence badge; our catalog lists it as MIT.
Langflow's pitch is source access: "Source code access lets you customize any component
using Python", and flows can be exported as JSON for Python apps. Flowise is a Node.js
server (server), a React UI (ui) and a components package of third-party node
integrations, so extending it is TypeScript work.
On the way out, both publish flows through an API, and both list MCP support: Langflow can serve flows as an MCP server and connect to external ones, and Flowise's docs list MCP client and server nodes. Langflow's README lists observability integrations with "LangSmith, LangFuse and other integrations". Flowise's docs list tracing, evaluations and human-in-the-loop features.
Flowise keeps its data in SQLite by default, or PostgreSQL, MySQL or MariaDB per our
catalog; our install uses Postgres and a bind-mounted data directory for credentials
and storage. Langflow's Docker docs state that it "stores its SQLite database inside the
container, not on a mounted volume", so removing the container deletes the data even if a
volume is mounted; set LANGFLOW_DATABASE_URL for PostgreSQL or
LANGFLOW_SAVE_DB_IN_CONFIG_DIR=true to keep SQLite on a persistent volume. The install on our
Langflow page mounts a volume on LANGFLOW_CONFIG_DIR, which per the same docs holds the
secret key and uploaded files; add LANGFLOW_SAVE_DB_IN_CONFIG_DIR=true (or a PostgreSQL
URL) so the database lands on that volume too.
In our install check (GCP e2-standard-2, Ubuntu 26.04), idle at first boot with the default install snippet and no flows running, Flowise used ~697 MB of RAM and ~9.5 GB of disk (November 2026), and Langflow used ~1.2 GB of RAM and ~3.5 GB of disk (September 2026). Those are idle numbers; a flow calling a model or building a vector index will use more. Flowise's disk figure covers its container image plus Postgres; we do not break it down further. Langflow's installation docs give a minimum of 2 GB RAM (4 GB recommended); the 1 GB on our Flowise page is our own estimate.
Both are 2 out of 5. Flowise: a Compose file with Postgres, port 3000, and a first
admin account registered in the browser. Langflow: one docker run on port 7860; the
image sets LANGFLOW_AUTO_LOGIN=false, and its README says to set
LANGFLOW_SUPERUSER_PASSWORD before starting, so you generate and keep a password.
Choose Langflow if you want a repository that has not been archived, Python components you can edit, and a plain MIT licence.
Choose Flowise if you already have flows built on it or you prefer its TypeScript stack and lighter idle RAM (~697 MB against ~1.2 GB), and you are willing to run a frozen release or maintain a fork.
Both installs are verified on Ubuntu 26.04; the pinned files are on the Flowise and Langflow pages. Keep model API keys out of exported flows, put a TLS reverse proxy in front before exposing either, and give Flowise more disk than the RAM figure suggests.
Flowise lists a lower minimum: 1024 MB vs 2048 MB for Langflow.
No. Flowise is Apache-2.0 core with commercially licensed enterprise modules, Langflow is MIT.
Langflow if you want a repository that is not archived and Python-editable components. Flowise only if you already depend on it and accept a frozen release or a fork. Flowise used less idle RAM (~697 MB against ~1.2 GB) but more disk (~9.5 GB against ~3.5 GB) in our first-boot check.
Its README opens with "Flowise has been archived" and GitHub shows the repository as read-only. The maintainers' announcement says they decided to wind down operations, new pull requests would no longer be reviewed, and that the Apache 2.0 code is yours to keep building on.
Yes: the Apache-2.0 code stays available and the install we verified still runs. Treat it as a frozen snapshot or fork it, since upstream has said it will not review new pull requests.
Not by default. Its Docker docs say it stores its SQLite database inside the container, not on a mounted volume. Set LANGFLOW_DATABASE_URL for PostgreSQL or LANGFLOW_SAVE_DB_IN_CONFIG_DIR=true to keep SQLite on a persistent volume.
Paid link — we earn a commission if you shop through it.
The team LLM-app platform vs. the Python-component canvas.
A whiteboard-and-docs workspace on a plain Compose file vs. native apps on a commercial server.
A full identity platform vs. a passkey-only sign-in box.