What is the difference between Dify, Flowise, Langflow and n8n?

Dify is a platform for LLM applications with workflows, knowledge base, agents and publishing as an app or API. Flowise was a visual builder for agents and LLM workflows (Assistant, Chatflow, Agentflow). Langflow is a Python framework with a visual editor for AI applications. n8n is a workflow automation platform with integrations and AI agents. The first three come from the AI field, n8n from classic process automation.

Is Flowise still active?

No. The repository is archived. According to the maintainers' announcement, a code freeze applied from 29 July 2026, the repository was archived on 13 August 2026 and the end of life on 31 August 2026 ended the core team's official presence on Discord and GitHub. Workday had acquired Flowise on 14 August 2025. New projects should not be built on Flowise.

No. n8n is under the Sustainable Use License and describes itself as fair-code, not open source. The source code is public, but use is restricted to internal business purposes and certain services. Files with .ee in the name are under the n8n Enterprise License. Until 17 March 2022, Apache 2.0 with Commons Clause applied.

Which license is Dify under, and is commercial use permitted?

Dify is under the Dify Open Source License, a modified Apache 2.0 license with two additional conditions: multi-tenant operation (one tenant corresponds to one workspace) requires written permission, and the logo and copyright in the frontend must not be removed. Commercial use is otherwise expressly permitted, including as a backend service for other applications. GitHub therefore shows the license as Other.

Which tool is best suited for RAG with your own documents?

Dify comes with a knowledge base with document import and a knowledge pipeline. Langflow offers embedding models, vector stores and knowledge bases as components. n8n provides vector stores and agents as nodes and is suitable when the documents are delivered from other systems. Which one fits depends less on the RAG function than on license, permissions and operations.

Do all four support MCP?

Yes, to varying degrees. Dify can import MCP server tools (HTTP transport only) and publish apps as MCP servers. Langflow works as an MCP client and as an MCP server. n8n has a built-in MCP server as well as an MCP client node and an MCP server trigger. Flowise had MCP client and server nodes.

Is SSO available in the free versions?

Mostly not. In Dify, SSO, multiple workspaces and custom roles belong to the Enterprise edition. Flowise offered OIDC only in the Enterprise edition. Langflow does not enforce role-based permissions in the open-source build and connects SSO through external authentication (a token from an upstream identity provider); its own OIDC, SAML and LDAP are available as a plugin. n8n provides SAML, OIDC and LDAP for self-hosting from the Business plan.

Which of the tools is the most secure?

That cannot be derived from the number of security advisories. In the twelve months to 21 October 2026, GitHub published 13 advisories for Dify, 109 for Flowise, 35 for Langflow and 217 for n8n. Many reports also show active review. What matters is a prompt update process, separated access and the question of who may execute code in operation.

May I host n8n or Dify for my customers?

That depends on the license and must be clarified with the vendor or legal counsel before starting. According to n8n's license FAQ, hosting n8n as a service is not permitted if customers build workflows on it; contract work and installation on the customer's server are permitted. With Dify, multi-tenant operation requires written permission. This article is not legal advice.

What replaces Flowise?

That depends on the use. For AI applications with a knowledge base and publishing as a chatbot or API, Dify is an option; for Python-oriented teams with an MIT license, Langflow; for process automation with AI steps, n8n. Flows cannot be transferred automatically; credentials, knowledge bases and prompts are carried over and the flow structure is rebuilt.

Dify vs Flowise vs Langflow vs n8n: Comparison 2026

Timo Wevelsiep22.10.2026

#Dify
#Flowise
#Langflow
#n8n
#AI
#Workflow
#SelfHosting
#License

Editorial note: The information in this article was compiled to the best of our knowledge at the time of publication. Technical details, prices, versions, licensing terms, and external content may change. Please verify the information provided independently, particularly before making business-critical or security-related decisions. This article does not replace individual professional, legal, or tax advice.

Have Dify, Langflow or n8n operated for you? WZ-IT takes over hosting and operations in Germany: Dify, Langflow and n8n, with updates, backups and monitoring. For existing Flowise installations, we assess the situation and plan the migration. Book an initial consultation

Anyone who wants to assemble AI applications rather than program them ends up with one of four tools: Dify, Flowise, Langflow or n8n. They look similar, but differ in origin, license and operating model. One of them, Flowise, is no longer being developed.

This article compares the four tools by the criteria that matter in operation: license and hosting for third parties, feature set, deployment, permissions and SSO, telemetry, security advisories and costs. All figures come from the GitHub API, license files and vendor documentation, as of 21 October 2026.

A note on scope: the tools in this article build workflows and agents. Anyone looking for a chat interface for employees compares Open WebUI, LibreChat and AnythingLLM instead, see Open WebUI vs. AnythingLLM.

Short answer
Flowise: archived and wound down
License and hosting for third parties
Feature comparison
Deployment and operations
Users, permissions and SSO
Telemetry and data flow
Security advisories of the last twelve months
Costs
Which tool for which use case
Common misconceptions
Data and methodology
How WZ-IT works
Further guides
Sources

Short answer

	Dify	Flowise	Langflow	n8n
Purpose	Platform for LLM applications with workflow, knowledge base and agents	Visual builder for agents and LLM workflows	Python framework with visual editor	Workflow automation with AI agents
License	Dify Open Source License (modified Apache 2.0)	Apache 2.0 plus commercial enterprise parts	MIT	Sustainable Use License plus Enterprise License
Open source per OSI	no	partly (core: Apache 2.0)	yes	no (fair-code)
Latest version	1.17.1 (10 Sep 2026)	[email protected] (29 Jul 2026)	1.12.7 (21 Oct 2026)	2.44.5 (21 Oct 2026)
GitHub stars	about 159,000	about 55,500	about 155,900	about 208,200
Status	active	archived, end of life	active, owned by IBM via DataStax	active
Main restriction	Multi-tenant and logo	discontinued	RBAC and own SSO only with plugin	Hosting for third parties, features by plan

Rule of thumb: Dify for AI applications with a knowledge base and publishing, Langflow for teams close to Python with the MIT license, n8n for process automation with AI steps. Do not introduce Flowise anew.

Flowise: archived and wound down

The maintainers documented the end of Flowise in an announcement and on the page flowiseai.com/sunset. The timeline:

Date	Event
14 Aug 2025	Workday acquires Flowise
29 Jul 2026	Code freeze according to the announcement, last version [email protected]
13 Aug 2026	Maintainers' announcement (discussion 6727), repository archived, issues and pull requests locked
31 Aug 2026	End of life, official presence of the core team on Discord and GitHub ends

The maintainers' rationale: coding agents are replacing rigid low-code workflows. The source code stays public, the Apache 2.0 part may be forked, the commercial enterprise parts may not. There is no statement on Flowise Cloud, enterprise contracts and security updates after the end of life. The website gives 10 August as the archiving date, the announcement and repository give 13 August; the last commit is from 13 August.

For operations, this means:

No new projects on Flowise.
Security situation: After archiving, GitHub published 20 further advisories, six of them from 10 September 2026 with no indication of a patched version. As no releases appear any more, fixes are not to be expected.
Startup failure in the Docker image: According to issue #6688, the Docker image 3.1.4 does not start (according to comments because of an unpinned dependency, connect-sqlite3); the issue is open and locked.
Existing installations: A migration should be planned while there is still time. Flows cannot be transferred automatically; credentials, knowledge bases and prompts are carried over, the flow structure is rebuilt.

License and hosting for third parties

The license matters more with these four tools than with most other open-source projects. Three of the four have a custom license, and GitHub shows NOASSERTION or Other there.

Tool	License	Additional terms
Dify	modified Apache 2.0	Multi-tenant operation only with written permission (one tenant corresponds to one workspace); logo and copyright in the `web/` frontend must remain; contributors agree that the vendor may make the license stricter or more relaxed
Flowise	Apache 2.0	Directory `packages/server/src/enterprise` under a commercial license (SSO, workspaces, RBAC)
Langflow	MIT	no additional terms
n8n	Sustainable Use License 1.0	Use for own internal business purposes; files with `.ee.` under the n8n Enterprise License, productive use only with a license

n8n. The vendor itself states that it does not call n8n open source. The license FAQ lists what is permitted: paid contract work (building, setting up, maintaining and training on workflows), multiple own instances, n8n as a backend in one's own product without end-user access to the workflow logic, and installation on a customer's server as long as one does not additionally host. Not permitted are, among other things, hosting n8n as a service where customers build workflows on it, enabling end users to build workflows (the FAQ also names MCP and AI agents as ways), forking n8n as an own automation product, or white-label offerings. The contributor agreement allows the vendor to change the license without asking. Until 17 March 2022, Apache 2.0 with Commons Clause applied.

Dify. Commercial use is permitted, including as a backend service for other applications. The two additional conditions concern platform operators: anyone who runs multiple tenants (workspaces) in one environment with the source code needs written permission, and the frontend logo must not be removed. The clause is tied to operating an environment with multiple workspaces. WZ-IT operates Dify as a dedicated single instance per customer with unchanged branding. The pricing page also describes the Community edition as "for non-commercial projects", while the license text permits commercial use.

Flowise and Langflow. Flowise started under MIT in April 2023, switched to Apache 2.0 in September 2023, added a multi-tenant and logo clause in November 2023, dropped it again in January 2024 and has carried commercial enterprise parts in the enterprise directory since May 2025. Langflow has been under MIT since its LICENSE file was added in February 2023; no license change is recorded.

This section classifies license texts and vendor notes and is not legal advice. Anyone operating one of the tools for third parties should clarify the license in advance with the vendor or legal counsel.

Feature comparison

All four cover the basic functions. The differences lie in the focus.

Feature	Dify	Flowise	Langflow	n8n
Focus	LLM apps	Agents and chatflows	Agents in Python	Process automation
Knowledge base and RAG	Knowledge base, knowledge pipeline, external knowledge bases via API	Document store, vector DBs as nodes	Embedding models, vector stores, knowledge bases	Vector stores and embeddings as nodes
Agents	Agent node, agents with own sandbox	Assistant, Chatflow, Agentflow (multi-agent)	Multi-agent orchestration	AI Agent node with tools and memory
MCP	Client (HTTP only) and server	Client and server nodes	Client and server	Server built in, client node, server trigger
Human-in-the-loop	"Human Input" node	Approval per tool or node	Pause and approve	Approval before tool call
Code	Python, JavaScript (sandbox)	custom code, expressions	Python in every component	JavaScript, Python
Triggers	Schedule, webhook, integrations	Prediction API	Webhook component, API	Webhook, schedule, queues, many apps
Publishing	Web app, API, embed, MCP	Embed widget, share, API	API, embed widget, MCP, `/api/v1/responses`	Chat Hub, webhooks, embedded chat
Integrations	Plugins and marketplace, number not stated	"100+" sources, tools, vector DBs	"Hundreds" of prebuilt components (website)	2,000+ per website (GitHub description says 400+)

The number of n8n integrations varies between 400, 500, 1,500 and over 2,000 depending on the source and cannot be compared reliably.

Deployment and operations

	Dify	Flowise	Langflow	n8n
Options	Docker Compose; Kubernetes officially only Enterprise	npm, Docker, Compose	`uv pip install`, Docker, Desktop, Helm charts	Docker, Compose, npm
Additional services	PostgreSQL, Redis, vector DB (default Weaviate), sandbox, worker: 7 core and 8 dependency containers	SQLite (default), MySQL, MariaDB, PostgreSQL; queue mode with Redis	SQLite (default), PostgreSQL; Redis for multiple workers	SQLite (default), PostgreSQL; queue mode with Redis
Scaling	Celery workers, multiple API replicas only with sticky sessions	Queue mode with any number of workers	multiple workers (`LANGFLOW_WORKERS`) with a Redis job queue	Queue mode with workers; multi-main only Enterprise

Dify is the heaviest package to operate: the Compose stack starts 15 permanently running containers (plus a one-time init container), and the documentation specifies at least 2 CPU cores and 4 GiB RAM. Langflow has official Helm charts (langflow-ide, langflow-runtime); for Dify, Kubernetes is officially reserved for the Enterprise edition. With n8n and Flowise, the entry point is a single container with SQLite; for queue mode, PostgreSQL and Redis are added.

Users, permissions and SSO

	Dify	Flowise	Langflow (OSS)	n8n
Roles	Owner, Admin, Editor, Normal	Workspaces with RBAC only Cloud and Enterprise	Superuser and further users; no RBAC enforcement	Owner, Admin, Member; project roles
Multiple workspaces	Enterprise only (Community: one workspace after installation; more via `ALLOW_CREATE_WORKSPACE`, then observe the multi-tenant clause)	Cloud and Enterprise only	no	Projects, assignment by edition documented contradictorily
SSO	Enterprise (Community: OAuth login with GitHub or Google)	OIDC, Enterprise only	external authentication (token from an upstream identity provider or OIDC proxy); own OIDC, SAML and LDAP as plugin	SAML, OIDC, LDAP from Business (self-hosting)
Audit log	not found in the documentation (run logs available)	login activity documented (Enterprise and Cloud); general audit log not found in the documentation	Audit table for authorization decisions (optional, off by default, retention 90 days)	Enterprise only
Versioning	Draft, live version and previous versions	not found in the documentation	Version history in the flow editor (save, preview, restore)	Workflow history (24 hours for all, full history Enterprise); Git and environments from Business

Langflow does not enforce role-based permissions in the open-source version: the documentation describes a pass-through service that allows every action; roles (viewer, developer, admin) are enforced only by a separate plugin. In addition, external authentication offers a coarse access ceiling taken from a JWT claim. Anyone operating Langflow for multiple teams therefore secures access via reverse proxy, network and identity provider.

According to its own statement, n8n delivers "almost the complete feature set" in the Community edition. Not included are, among others, variables, environments, external secret stores, log streaming, multi-main, SSO and Git integration.

Telemetry and data flow

Tool	Default	Disabling
Dify	Update check to `updates.dify.ai`, Next.js and Weaviate telemetry in the Compose stack; own product telemetry not documented	Empty the update URL, disable Next.js telemetry and set `WEAVIATE_DISABLE_TELEMETRY=true`
Flowise	off by default since May 2025; sends data only if `POSTHOG_PUBLIC_API_KEY` is set	`DISABLE_FLOWISE_TELEMETRY` has no effect in current versions
Langflow	anonymous telemetry on (version, platform, flow runs and component executions with duration, success and error message); Desktop additionally sends the registered email address	`DO_NOT_TRACK=True`
n8n	Telemetry on (events, execution counts, instance pulse every six hours), no workflow data	`N8N_DIAGNOSTICS_ENABLED=false`; Business and Enterprise license keys ping the license server daily

For environments with network isolation, this is a check before rollout: disable telemetry and update checks and limit outbound connections with a firewall.

Security advisories of the last twelve months

Published GitHub security advisories from 21 October 2025 to 21 October 2026:

Tool	Advisories	of which critical	of which high
Dify	13	0	6
Flowise	109	28	63
Langflow	35	16	12
n8n	217	22	97

The number says nothing about the quality of a project. It shows how much was reported and published and depends on size, attack surface and reporting culture. The patterns stand out: for n8n, the critical reports mostly concerned remote code execution (among others via Merge, Git, XML and Code nodes) and sandbox escapes; since June 2026 no critical reports have appeared, and all entries have a patched version. For Flowise, 20 reports fell in the period after archiving.

The same applies to all four: workflow builders execute code and hold credentials for many systems. Apply updates promptly, restrict access and do not expose the instance directly to the internet.

Costs

Self-hosted, there is no license fee as long as the license terms are met and no enterprise features are used. The vendors additionally offer cloud plans (pricing pages, as of 21 October 2026; Dify excluding taxes, n8n without a net or gross statement):

Tool	Cloud plans	Enterprise
Dify	Professional USD 59 per workspace and month (USD 590 annually), Team USD 159 (USD 1,590 annually)	on request
Flowise	Free USD 0, Starter USD 35, Pro USD 65 per month; page carries the notice about discontinuation	on request, no statement after the end of life
Langflow	no prices published	Contact page for professional services and premier support
n8n	Starter EUR 20, Pro EUR 50 per month (annual billing); Business EUR 667 per month, currently self-hosting only	on request

Operating costs arise from hosting, updates, backups, monitoring and maintaining the flows.

Which tool for which use case

Use case	Best fit	Points to note
AI applications with a knowledge base, publishing as a chatbot or API	Dify	Multi-tenant and logo clause; SSO and roles only Enterprise; operation with 15 containers
Teams close to Python, flows as traceable code, MIT license	Langflow	No RBAC enforcement in the OSS build; secure access yourself; weekly releases
Process automation with AI steps, many third-party systems	n8n	fair-code license; hosting for third parties; features by plan
Existing Flowise installation	Plan the migration	Target tool by use case; rebuild flows
Chat interface for employees	Open WebUI, LibreChat or AnythingLLM	different tool class, see comparison

The tools can be combined: n8n can call a Dify workflow or a Langflow flow via its API and thus integrate an AI step into a process.

Common misconceptions

"n8n is open source." No. n8n calls itself fair-code. The code is viewable, use is limited by the Sustainable Use License.
"Dify is under Apache 2.0." Not unmodified. A modified Apache 2.0 license applies, with a multi-tenant and logo condition.
"Flowise is a safe choice because it is so widespread." The project is archived and has been at end of life since 31 August 2026.
"Langflow has roles and permissions." The concepts are documented but not enforced in the open-source build.
"Self-hosted means no data leaves the house." With Dify, Langflow and n8n, the update check or telemetry, respectively, is active by default and must be disabled for isolated environments.

Data and methodology

As of: 21 October 2026. The data basis is the GitHub API (license, version, stars, archive status, published security advisories), the repositories' LICENSE files and the vendors' documentation. Stars are rounded and change daily. The advisories cover the period from 21 October 2025 to 21 October 2026; if a patched version is missing in the metadata, this does not necessarily mean that no fix exists. Pricing pages carry no date. Where documentation and pricing page contradict each other (n8n: projects and queue mode), this is noted.

Tables and figures may be reused with attribution and a link to this article.

How WZ-IT works

WZ-IT operates Dify, Langflow and n8n as a hosting and operations offering in Germany: updates, backups, monitoring and the connection to models, whether European providers via LLM hosting or own inference. Before operating for third parties, we classify the license situation of the respective tool with you; we do not provide legal advice. For existing Flowise installations, we take stock of the setup, choose the target tool with you and carry over credentials, knowledge bases and prompts; the flows are rebuilt in the target tool. The assessment of Flowise installations describes the process.

Further guides

Open WebUI vs. AnythingLLM, the comparison of chat interfaces.
AI agent frameworks compared, n8n, LangGraph, CrewAI and more.
Open-source license changes, the chronicle and what operators must do.
Open WebUI license and branding, an example of additional clauses.
AI platforms for businesses compared, hosted offerings by GDPR and Section 203.
LLM hosting, operating models in Germany.

Choose a workflow builder or have it operated? We classify use case, license and permissions and operate Dify, Langflow and n8n in Germany. For Flowise, we plan the migration. Book an appointment