Dify is a platform for building production-ready AI agents, workflows, knowledge pipelines and tools. The page describes it as one canvas where teams can build agents, prepare knowledge bases, choose models and publish the result.
The workflow editor lets teams define how an app uses prompts, retrieves data, makes decisions, calls tools, asks a person for input and completes tasks. The logic stays visible so it can be tested, debugged and handed over. Agents can be built by chatting or by configuring them by hand, then used as an app, embedded in a website or reused as a node inside a workflow.
A knowledge pipeline handles files, websites, online documents and drives. It extracts, cleans, chunks and indexes content, and lets users test retrieval before the knowledge base is connected to agents or workflows. A Marketplace provides model providers, tools, data sources and MCP integrations, and approved plugins can be reused across apps.
Finished apps can be published as hosted experiences, API endpoints, embeds or MCP-compatible tools. Logs, feedback, annotations, latency and usage data help teams see what happened and improve the app over time.
Dify is aimed at businesses, from small teams building their first production AI apps to enterprises that need access control and audit logs. It is offered as Dify Cloud, a managed service, as Dify Enterprise for self-hosting or running in a private VPC, and as Community Edition, open source and self-hosted with Docker.
Dify Cloud has a free Sandbox plan with 200 message credits, one workspace and one team member. The Professional plan is listed at $59 per workspace per month, or $590 per workspace per year when billed annually. The Team plan is listed at $159 per workspace per month, or $1,590 per workspace per year. Enterprise pricing is by contact with sales.
