Dify

Dify is a platform for building and deploying agentic workflows, agents, and retrieval-augmented generation pipelines on one visual canvas. Teams define prompts, knowledge retrieval, branching, tools, and human approval steps as visible building blocks, then publish the result as a web app or an API. Dify runs as a managed cloud service, inside a private VPC, or as a self-hosted open-source community edition.

Key Features

• Visual workflow studio for prompts, retrieval, branching, tools, and human input.
• Agent runtime with plugins, tool calls, and a knowledge pipeline for documents.
• Model-agnostic setup that connects hosted or self-hosted model providers.
• Publish apps as web pages or APIs, with logs, annotations, and monitoring.
• Self-hosted Community Edition with a single-command Docker deployment.
• Enterprise controls including SSO and SAML, role-based access, and audit logs.
• Reusable templates and a plugin marketplace for common app patterns.

Who is it for

• Developers and AI engineers building RAG applications, agents, and internal tools.
• Product teams that want to prototype an AI app and ship it without rebuilding the stack.
• Enterprises that need a private deployment, audit trails, and role-based access.
• Operations teams turning a repeated process into a guided AI workflow.

What can you do with Dify

• Build a knowledge assistant that answers from your own documents.
• Design a multi-step agent workflow with tools, conditions, and human review.
• Compare model providers inside one app and switch without redesigning the flow.
• Publish a web app or API endpoint and monitor how it is used.
• Self-host the community edition or deploy in a VPC for private data.
• Reuse a template for support, research, or content workflows.

How does Dify work

1. Create a Dify Cloud workspace or deploy the community edition yourself.
2. Add a model provider and store the credentials in the workspace.
3. Build an app on the canvas with prompts, knowledge retrieval, tools, and branches.
4. Test the flow with sample inputs and inspect each step in the logs.
5. Publish the app as a web page or API and monitor usage.
6. Upgrade to an enterprise deployment when you need SSO, audit logs, or a VPC.

Pricing

Dify Cloud lists a free Sandbox plan, Professional at $59 per workspace per month or $590 per workspace per year, Team at $159 per workspace per month or $1,590 per workspace per year, and Enterprise with custom pricing. The self-hosted Community Edition is free under its source-available license.

FAQ

Is Dify open source?

The Community Edition is source-available and can be self-hosted with Docker under Dify's license. The cloud service and enterprise edition add managed infrastructure and commercial features.

What can I build with Dify?

You can build chatbots, retrieval-augmented knowledge assistants, multi-step agent workflows, and internal tools that combine prompts, tools, and your own documents.

Does Dify work with different model providers?

Yes. Dify connects to a range of hosted and self-hosted model providers, so an app can use different models for reasoning, embedding, or speech without changing its structure.

Can I self-host Dify?

Yes. The community edition deploys with Docker, and the enterprise edition supports VPC and on-premises installations with SSO, audit logs, and support.

Is there a free plan?

Dify Cloud includes a free Sandbox plan with limited credits and app capacity, and the self-hosted Community Edition is free to run under its license.

How do I move an app to production?

Publish the workflow as a web app or API after testing it, then monitor logs and usage. Enterprise deployments add private networking, governance, and support.