Dify
An open-source, visual builder for AI apps and agent workflows, with a knowledge base you can query behind them.
What it costs
| Tier | Price | For | Includes |
|---|---|---|---|
| Sandbox | $0 | Trying Dify's core features before committing to a workspace. | 200 message credits, 1 team workspace, 1 member, 5 apps, standard document processing. |
| Professional | $590/year per workspace | Independent developers and small teams shipping a production AI app. | 5,000 message credits/month, 3 team members, 50 apps, priority document processing. |
| Team | $1,590/year per workspace | Medium-sized teams needing collaboration and higher throughput. | 10,000 message credits/month, 50 team members, 200 apps, top-priority document processing. |
Read from Dify's own pricing page on 25 Sep 2026. Vendors change prices without notice.
Where it breaks
Message credits cover trying different models (OpenAI, Anthropic, Gemini, xAI, Tongyi) inside Dify; once they're used up you switch to your own API key, which is billed separately by the model provider.
Available as Dify Cloud (hosted) or self-deployed via Docker as the free, open-source Community Edition. A separate Dify Enterprise tier covers private deployment, priced on request.
First entry in this reference. There is no previous check to compare it against.
Dify sits a level below the chat-assistant tools in this directory: it’s for building the thing, not using someone else’s finished version of it. A visual canvas lets you wire together prompts, a knowledge base, and tool calls into an agent or workflow, without writing the orchestration code by hand.
The Community Edition — free, self-hosted, open-source — is why Dify shows up so often in engineering teams’ stacks: you can run it on your own infrastructure with no per-seat cost, at the price of managing the server yourself. Dify Cloud removes that operational burden for a yearly fee per workspace, with message credits standing in for the model calls until you bring your own API key.
This is a builder’s tool, not a business user’s. Getting value from it assumes someone on the team is comfortable thinking in workflows and knowledge bases, and it competes less with ChatGPT than with Make or a custom Python script — the question is whether the workflow needs LLM reasoning at each step or just data movement.
What it does not give you is a finished agent. You still have to design what yours should do.
Get Dify
Buying the tool is not the hard part.
The ebook is about what to hand over, in what order, and how you find out when it goes wrong. One payment, no subscription.