Flowise Forks Multiply After Archive: Flow-Wiser vs Keelflow

What happened

Flowise, the visual builder that let developers wire LangChain agents together without much code, has officially reached the end of its road. Its maintainers froze feature work on July 29, 2026, archived the GitHub repository on August 13, and set August 31 as end of life 4. At that point the core team's presence on Discord and GitHub was scheduled to end 2. The project had more than 55,000 GitHub stars and an Apache 2.0 license when it was archived 4.

The exact dates vary slightly depending on who you ask. The Flow-Wiser fork lists an end-of-life announcement on August 3, a code freeze on July 29, and an upstream archive date of August 10 1. Another account notes that GitHub's banner shows August 13, three days after the planned date 2. The discrepancies are minor, but they show how quickly the record of a wind-down can blur.

The shutdown notice did not just say goodbye. It pointed out that the Apache 2.0 code would stay on GitHub and invited teams to fork it 2. Within weeks that invitation produced several competing continuations.

The forks arriving to fill the gap

The GitHub topic page for "flowise" now lists 92 public repositories 3. Two of them stand out because they present themselves as successors rather than personal snapshots.

Flow-Wiser, maintained under the dblagbro account, calls itself an unofficial community continuation fork. It explicitly disclaims any affiliation with FlowiseAI, Inc. or Workday 1. Its main pitch is licensing. Flow-Wiser says it is the first Flowise container anyone may freely redistribute 1. To get there, it removed commercially licensed files and rebuilt authentication, RBAC, SSO, MFA, audit logging, encryption at rest and multi-tenancy from scratch under Apache 2.0. The project says this was done without reading the original files 1. Its build is also designed to fail if any commercially licensed artifact shows up in the image 1. The current release is tagged 3.1.4-fw10 1.

Keelflow, from the Perruer account, takes a narrower approach. It describes itself as a security-maintained continuation offering security fixes, an Apache-2.0-only codebase, and drop-in compatibility with existing flows and data 3. Both projects showed activity in early October, with Keelflow updated October 8 and Flow-Wiser October 10 3.

The two share a core promise: a cleanly Apache-licensed Flowise. Where they differ is ambition. Flow-Wiser is re-engineering enterprise features. Keelflow is pitching continuity and patching.

Why the licensing angle matters

The emphasis on licensing is not just pedantry. Flow-Wiser's claim that it is the first freely redistributable container suggests that the official Flowise images contained commercially licensed components alongside the Apache code 1. Anyone wanting a fully open build therefore has to cut those parts out and replace them.

That matters more because the obvious alternatives have their own licensing complications. One survey of open-source agent tooling warns that n8n, Dify and the AutoGPT platform all carry commercial restrictions that fall outside the open source definition. It advises teams to read the LICENSE file before committing 4. For organizations that picked Flowise specifically for its permissive license, a straight move to n8n or Dify may not be a like-for-like replacement.

The case against forking

Not everyone is convinced the forks are the answer. One practitioner who has stopped starting new projects on Flowise notes that the repository has roughly 25,000 forks. The same practitioner cautions that any fork makes you responsible for security patches, dependency upgrades and every future breaking change in LangChain 2. That writer compares Flowise against n8n, which has about 206,800 stars and released version 2.42.4 on October 7, and Dify, which has about 158,000 stars and released 1.17.1 on September 10 2. Both are clearly active projects with momentum.

There are also the code-first frameworks. LangChain, LangGraph, CrewAI, Google's ADK, Pydantic AI and others have shipped releases in recent weeks under MIT or Apache 2.0 licenses 4. The same survey stresses that free licensing does not mean free operation. Engineering, inference, hosting, patching and tracing still cost money 4.

Reading the situation

The "fork it yourself" advice is doing what it was supposed to do, but it is fragmenting the effort along the way. With dozens of repositories and at least two serious successor bids, Flowise users now have to judge which fork has the staying power to keep up with LangChain's pace. That is the same maintenance burden the original team chose to drop.

My reading is that these forks are best seen as a lifeline for existing deployments rather than a foundation for new projects. Keelflow's drop-in, security-first pitch fits teams that need to keep current flows running while they plan a migration. Flow-Wiser's clean-room enterprise rebuild is more ambitious and potentially more valuable, but it also carries more risk. Its long-term value depends on whether contributors gather around it. Historically, community forks survive when they consolidate around one effort, not when they splinter. If one of these two can draw in the bulk of Flowise's former contributors, Flowise may get a real second life. If not, the 25,000 forks may end up as snapshots of a tool people eventually move away from.