The best Flowise alternatives, compared honestly
Flowise announced its sunset on 29 July 2026 and reaches end of life on 31 August. The repository is already archived, version 3.1.4 shipped the same day as the notice, and 1,047 open issues were frozen where they stood. The team handled it well: a month of warning, a plain reason, and the code left standing. This page is not an obituary, because your self-hosted instance does not stop working. It is an inventory of what the sunset actually removes, including two things the notice does not spell out, and then 7 alternatives ranked by the only question that matters after a shutdown: could the owner do this to you again, and would the open-source core survive it? Every fact was verified against the licence files, the GitHub API and vendor pricing pages on 20 August 2026.
Nothing here reads a Flowise chatflow, so every route is a rebuild. Pick by what you are optimising for:
- Closest replacement, and it has already survived this: Langflow, clean MIT, nothing carved out, a release most weeks.
- Biggest ecosystem: Dify, but read the licence first, because it forbids running Dify as a multi-tenant service.
- Most permissive licence: Sim, plain Apache-2.0 with no enterprise directory anywhere in the tree.
- Before you pick anything: if you run Flowise enterprise SSO or RBAC, those parts were never Apache 2.0, and the subscription that licensed them is no longer purchasable.
The event
What the sunset actually does
Straight from flowiseai.com/sunset, signed by Henry and the Flowise team. Note the heading the vendor put on its own milestone table: "Impact on OSS Repository".
| Date | Milestone | What it means in practice |
|---|---|---|
| 29 July 2026 | Announcement and code freeze | "Active feature development ceases immediately. We will no longer be reviewing or accepting new Pull Requests." Release 3.1.4 shipped the same day and is the last one. |
| 10 August 2026 | Repository archival | "The code will remain visible indefinitely, but issues and PRs will be locked and npm packages and Docker images will be marked deprecated." 1,047 open issues froze in place. |
| 31 August 2026 | End of life | "Official core team presence in Discord and GitHub will conclude." Nothing switches off; the people simply leave. |
The stated reason, in full and worth reading rather than paraphrasing: "As AI models become more capable at reasoning, we've noticed that developers are increasingly relying on new coding agents such as Claude Code/OpenClaw to handle complex tasks. The typical rigid workflow low code approach quickly hits the limit when it comes to complexity." That is a claim about the category, not an excuse, and if you agree with it then the right replacement may not be another canvas at all. Workday acquired Flowise in August 2025 and said publicly that Flowise was not going anywhere, so the commercial picture clearly changed as well. Both can be true at once.
What does not change
Your instance keeps running. There is no licence key, no phone-home and no kill switch, so 3.1.4 will behave on 1 September exactly as it did on 31 August. The Apache 2.0 code stays public and stays yours. Anyone telling you to migrate this week because the software expires is wrong.
What does change, and faster than usual
Maintenance ends. For most software that buys you a quiet year. For a tool whose whole job is calling model providers, it is sharper: provider APIs, model names and SDK versions churn every few weeks, and from now on every one of those breakages is yours to patch. The same applies to CVEs in a dependency tree this size.
Not everything was Apache 2.0
The notice says "the Apache 2.0 licensed code is yours to keep building on". The repository's own LICENSE.md is narrower: everything under packages/server/src/enterprise, plus marked files such as IdentityManager.ts, sits under a separate Commercial License. That directory holds sso/ and rbac/. The commercial terms permit production use only with a valid Enterprise Edition subscription, and that subscription is going away with the company. GitHub's licence detector reads the repo as NOASSERTION, not Apache-2.0.
"Fork it" has not happened yet
Three weeks after the team encouraged teams to fork, there is no community successor. Sorted by stars, the largest of 24,932 forks has 15, and none of the leading forks had been pushed since the 29 July announcement. A fork count measures how many people once clicked a button. Check the list yourself before planning around a rescue.
The notice covers the repo, not the cloud
The milestone table is headed "Impact on OSS Repository" and stops there. There is no shutdown date for Flowise Cloud, no data-export deadline and no refund policy anywhere in the announcement, and the docs site carried no end-of-life notice when checked. If you pay for hosted Flowise, that silence is your cue to export now and ask billing directly.
Pin your images today
npm packages and Docker images are marked deprecated, not deleted, but deprecated artefacts are exactly what disappears quietly two years later. Pull the image you are running, push it to your own registry, and pin the digest rather than the tag. This costs ten minutes now and is unrecoverable if you leave it.
The shortlist
7 Flowise alternatives worth evaluating
After a shutdown the ranking question changes. Feature lists matter less than who owns the project, whether the licence covers the whole product, and whether releases are still landing. Those are the three things that decide whether you are reading this page again in eighteen months about a different tool. Star counts and release dates below were read from the GitHub API on 20 August 2026; licences were read from the repositories, not from vendor marketing.
The same shape of product as Flowise, a visual canvas for chaining models, tools and retrieval, and the reason it ranks first is not its 153,497 stars. It is that Langflow already went through the event that just ended Flowise. After IBM acquired DataStax, the hosted DataStax Langflow was deprecated on 9 March 2026 and shut down on 9 April 2026, with DataStax Langflow data deleted. The open-source project came through intact under IBM stewardship and now ships a release most weeks: v1.11.4 on 19 August, v1.11.3 on 11 August, v1.11.2 on 4 August. The licence explains the difference: plain MIT, no enterprise directory, no carve-out, no clause about multi-tenancy, so when the commercial layer was withdrawn there was still a complete product underneath. Weakness: the corporate owner is now IBM, so this is stewardship rather than community governance, and self-hosting is the primary deployment path since the managed option went away.
Comparable in scale at roughly 153,000 stars, broader than Flowise ever was (agents, RAG pipelines, datasets, observability, an app runtime), and the Community Edition is genuinely free to self-host. The catch is written into the licence file rather than hidden: Dify is "licensed under a modified version of the Apache License 2.0, with the following additional conditions", and those conditions forbid using the source to operate a multi-tenant environment without written authorisation and forbid removing or modifying the logo and copyright information in the console. Fine for internal use, a blocker if you intended to resell it. Cloud pricing is per workspace rather than per seat: Sandbox free (1 member, 5 apps, 200 message credits, 50 knowledge documents), Professional $590 a year with 5,000 monthly credits and 3 members, Team $1,590 a year with 10,000 credits and 50 members. SSO sits in an unpriced Enterprise tier. Latest release 1.16.1 on 28 July.
The newest serious entrant and the fastest-moving: 29,438 stars and v0.8.7 released on 20 August 2026, the day this page was written. Its licence is the reason it outranks bigger names: plain Apache-2.0, detected as such by GitHub, with no enterprise directory and no additional conditions, which is precisely the property Flowise turned out to lack. Cloud is credit-metered: Free 1,000 monthly credits with 10 concurrent executions and 5 GB storage, Pro $25 per user per month billed monthly with 6,000 credits, Max $100 per user per month with 25,000. One honest wrinkle to read correctly: the pricing table marks self-hosting as Enterprise-only, but Apache-2.0 grants the right to self-host regardless of how it is packaged commercially; what Enterprise buys is support, not permission. Weakness: youngest project here, so the smallest ecosystem of connectors and the shortest track record.
The pragmatic pick, and the one most commenters reached for when the sunset landed. At 201,272 stars and 2.35.5 shipped on 20 August, it is the healthiest project on this page by a distance, with a connector library nothing here approaches, and it has grown genuinely capable agent nodes rather than only deterministic automation. Say the licence plainly though: n8n is under the Sustainable Use License, which is source-available and not OSI-approved, and files with .ee. in the name or path are excluded from even that and require an enterprise licence. Billing is the friendliest unit in the category: one execution is one charge however many steps it contains. EUR 20 a month annual for 2,500 executions on Starter, EUR 50 for 10,000 on Pro, EUR 667 for 40,000 on Business, unlimited users and workflows on every plan. Weakness: it is an automation platform that learned AI, so the LLM-native pieces (retrieval, evaluation, prompt management) are shallower than in Langflow or Dify.
23,921 stars, 0.88.3 on 19 August, and an open-core split that is at least honest about itself: MIT everywhere except packages/ee/, which has its own licence file, so you can see the boundary by reading the tree rather than by hunting for copyright headers in individual files. Same billing logic as n8n: running a flow costs 1 credit however many steps are in it, with AI priced by model class (fast 2, smart 10, frontier 20, or 1 if you bring your own key) and overage at $0.007 per credit. Free forever with no card, Plus $16 a month billed yearly for up to 5 users, Team $166 a month for 25. Weakness, and it is the usual one: the Community Edition has no cap on runs, users or flows, but excludes Agents and Chat, Projects, API access, SSO, user roles, audit logs, secret managers, branding and Git Sync, which is a lot of the reason a team would deploy it.
Narrower than the rest: a conversational-agent platform rather than a general workflow canvas, so it fits if what you built in Flowise was a chatbot over your own documents, and fits badly otherwise. The repository is MIT with no carve-out found, at 14,872 stars and pushed on 19 August, which is a better licence position than most of this list. The gap is commercial transparency: the pricing page returns HTTP 403 to non-browser requests, so no figure appears in the table below. That empty cell is a finding rather than missing homework, and it is the same pattern worth noticing anywhere: a vendor that will not serve its price to a script is a vendor whose price you cannot monitor over time.
A genuinely nice thing, and it is here to make one column impossible to ignore. Rivet is a desktop IDE and TypeScript library for building agent graphs, MIT with no carve-out, built and open-sourced by Ironclad, and the local-first model means there is no cloud that can be switched off. But the numbers say what the marketing will not: 4,670 stars, and the last release, app-v1.11.3, was published on 8 August 2025, more than a year before this page. The repository saw a push in July 2026, so it is not abandoned, yet a year without a release from a single-vendor project is precisely the state Flowise was in before its notice. Pick it with open eyes for a local, code-adjacent workflow, and do not pick it as the thing your production stack depends on.
The option the Flowise team effectively recommended: no canvas at all
Take the stated reason seriously, because it is a claim about your workload and not only about theirs. If the thing that pushed you to Flowise was gluing a few model calls together with retrieval and a branch or two, then a canvas was always the packaging rather than the value, and that logic fits comfortably in a file in your own repository using a framework such as LangGraph, the Vercel AI SDK, or nothing at all. What you gain is the stack you already know: version control, code review, tests, and a dependency you can pin. What you lose is real and worth weighing, because it is why visual builders existed in the first place: non-engineers can no longer read or change the flow, and you own the runtime, the queue and the observability yourself. Rebuilding a Flowise chatflow in code is often a day of work. Rebuilding it on another canvas is also a day of work. Since neither path preserves your existing graph, this is the rare migration where the cheapest option and the most durable one might be the same one.
Side by side
Flowise alternatives compared
The columns are chosen to answer the question a shutdown creates: could this happen again here, and what is left standing if it does? Licences were read from each repository's own licence file. Stars and release dates come from the GitHub API on 20 August 2026. Currencies are as published and never converted.
| Tool | Licence | Carved out of it | Who owns it | Stars | Last release | Free to self-host | Cloud entry price |
|---|---|---|---|---|---|---|---|
| Flowise (archived) | Apache 2.0, split | Yes: enterprise/ incl. sso/, rbac/ |
Workday, since Aug 2025 | 55,383 | 3.1.4, 29 Jul 2026 (final) | ✓ core only | no future |
| Langflow | MIT | ✓ nothing | IBM, via DataStax | 153,497 | v1.11.4, 19 Aug 2026 | ✓ | hosted option withdrawn Apr 2026 |
| Dify | Apache 2.0, modified | Yes: no multi-tenant use, no logo removal | LangGenius | 153,000 | 1.16.1, 28 Jul 2026 | ✓ single workspace | $590/yr per workspace |
| Sim | Apache-2.0, clean | ✓ nothing | Sim | 29,438 | v0.8.7, 20 Aug 2026 | ✓ by licence; marketed as Enterprise | $25 per user/mo |
| n8n | Sustainable Use License, not OSI | Yes: .ee. files need an enterprise licence |
n8n GmbH | 201,272 | 2.35.5, 20 Aug 2026 | ✓ community edition | EUR 20/mo, 2,500 executions |
| Activepieces | MIT | Yes: packages/ee/ |
Activepieces Inc | 23,921 | 0.88.3, 19 Aug 2026 | ✓ no run, user or flow cap | free, then $16/mo yearly |
| Botpress | MIT | ✓ none found | Botpress Inc | 14,872 | pushed 19 Aug 2026 | ✓ | pricing page returns 403 |
| Rivet | MIT | ✓ nothing | Ironclad Inc | 4,670 | app-v1.11.3, 8 Aug 2025 | ✓ local-first, no cloud | none, desktop app |
Read the third column before the fifth. Four of the seven have something carved out of the licence, and Flowise is on this page because the carved-out part was the part you cannot replace yourself when the vendor leaves. The two rows with nothing carved out and a release inside the last week, Langflow and Sim, are the two that would survive their owner losing interest. Note also that popularity predicted nothing here: Flowise had 55,383 stars the day it froze. Two columns were deliberately left out: connector counts, which every vendor counts differently, and benchmark-style feature grids, which go stale faster than this page can be updated. Verified 20 August 2026; check each project before committing.
Primary sources: Flowise sunset notice · Flowise LICENSE.md · Langflow · DataStax Langflow docs · Dify pricing · Sim pricing · n8n pricing · Activepieces pricing · Rivet
A fair call
Do you actually need to move yet?
The loud advice this month is to migrate immediately. That is right for some deployments and wrong for others, and the difference is not how much you like Flowise. It is which parts of it you are running and who is exposed when something breaks.
You can take your time if…
- You self-host a pinned version behind your own authentication. Nothing expires, no server phones home, and the risk accumulates gradually rather than arriving on 31 August.
- You bring your own model keys and use mainstream providers. Your breakage clock is set by how fast those APIs change, which is measured in months, not days.
- The flow is internal and a human notices when it fails. An internal tool that breaks in November and is fixed in November is an inconvenience, not an incident.
- You do all of this first, this week: export every flow as JSON, pull your Docker image into your own registry and pin it by digest, and record which Flowise version you are actually on. Ten minutes each, and unrecoverable if the artefacts are pruned later.
Move now if…
- You use Flowise enterprise SSO or role-based access control. Those live under the commercial licence, production use requires a subscription, and the subscription is going away with the company. This is the one genuinely urgent case.
- You are on Flowise Cloud. The sunset notice covers only the open-source repository and names no cloud date, export deadline or refund policy, and the docs carry no notice either. Export today and put the billing question in writing.
- Anything customer-facing runs through it. From now on an unpatched dependency is your incident and your disclosure, with no upstream advisory to wait for.
- You were about to build something new on it. Whatever the merits, starting a project on an archived repository in August 2026 is a decision you will be asked to justify later.
- Your plan was that the community would fork it. As of 20 August the largest fork has 15 stars and none has been pushed since the announcement, so that plan currently has nothing behind it.
Common questions
Common questions about the Flowise shutdown
Does my self-hosted Flowise stop working on 31 August 2026?
No. Nothing in the sunset notice disables running instances, and there is no licence key, no phone-home and no kill switch. Version 3.1.4 will keep doing on 1 September exactly what it did on 31 August. What ends is maintenance: no more pull requests, no more releases, locked issues, and npm packages and Docker images marked deprecated. For most software that would buy you a year of quiet. For a tool whose entire job is calling model providers, it is sharper than it sounds, because provider APIs, model names and SDK versions churn every few weeks and nobody upstream will be tracking them for you.
Is all of Flowise really Apache 2.0, as the sunset notice says?
Most of it, and this is the detail worth checking before you commit to a fork. The repository LICENSE.md states that everything under packages/server/src/enterprise, plus individual files carrying an explicit copyright notice such as IdentityManager.ts, is licensed under a separate Commercial License, and that only the content outside those paths is Apache 2.0. That enterprise directory contains the sso and rbac folders. The commercial licence says the code may only be used in production if you hold a valid FlowiseAI Enterprise Edition subscription, and the company that sells that subscription is winding down. GitHub's own licence detector reports the repository as NOASSERTION rather than Apache-2.0, which is consistent with a split licence.
Has anyone forked Flowise to keep it alive?
Not meaningfully, as of 20 August 2026. The repository has 24,932 forks, but sorted by stars the largest has 15, and none of the leading forks had been pushed since the announcement on 29 July. A fork count is a measure of how many people once clicked a button, not of whether a successor project exists. If you are counting on the community to carry Flowise forward, check the fork list yourself before you plan around it, because at the time of writing there was no community-led successor to point at.
Why did Flowise shut down when it had 55,000 stars?
The team's own reason, from the sunset page, is that the category is being squeezed from above: as models get better at reasoning, developers increasingly hand complex work to coding agents, and a rigid low-code workflow hits its ceiling as complexity grows. Workday acquired Flowise in August 2025 and said at the time that Flowise was not going anywhere, so the commercial context clearly changed too. Both things can be true. Stars measure how many people liked the idea in 2023 and 2024, not how many were paying for it in 2026.
What is the closest replacement for Flowise?
Langflow, on the merits rather than on popularity. It is the same shape of product, a visual canvas for chaining models, tools and retrieval, it is clean MIT with nothing carved out, and it has already survived the exact event that just killed Flowise: the hosted DataStax Langflow was deprecated on 9 March 2026 and shut down on 9 April 2026 with its data deleted after IBM acquired DataStax, and the open-source project came through whole and now ships a release most weeks. If you want the biggest ecosystem instead, Dify is comparable in size but its licence bans running it as a multi-tenant service. If you want something that never depended on a canvas, the honest answer for many teams is to move the logic into code.
Can I export my Flowise flows into another tool?
You can export them out of Flowise as JSON, and that is worth doing now rather than later, but no alternative on this page imports a Flowise chatflow. The node graphs are tool-specific, so the JSON is documentation of what you built rather than a portable artefact. Treat migration as rebuilding the flow on the new canvas with the old one open beside it. The pieces that do carry over cleanly are the parts that were never Flowise-specific: your prompts, your system messages, your retrieval corpus and your API credentials, which are not included in the export and must be recreated by hand.
Should I move off Flowise immediately?
Not necessarily, and the honest advice depends on what you are running. A pinned, self-hosted instance doing one internal job behind your own auth can reasonably keep running for months while you plan, because it will not break by itself. Move with urgency in three cases: you use the enterprise single sign-on or role-based access control, which are the parts under the commercial licence; you depend on Flowise Cloud, which the sunset notice never mentions since its milestone table covers only the open-source repository and gives no cloud date, export deadline or refund policy; or the deployment is customer-facing, where an unpatched dependency becomes your incident rather than an upstream one.