Dify

The trait that decides it is who else can use it. A product manager can open the studio, change a prompt, add a step and see the result, without waiting for an engineer or opening an editor. That shifts where iteration happens in an organisation, and it is worth more than any single feature in the list.

Pick it when subject-matter experts should own the prompts and the retrieval. Pick n8n when the work is mostly moving data between systems, because this is built around language applications first and integrations second.

The risk is that changes become invisible. Logic assembled visually lives in the platform's own storage, not in files a reviewer reads, so the question of who altered a production prompt last Tuesday has no answer of the kind engineering has relied on for thirty years. Rollback and staging become platform features rather than properties of your existing process.

Decide how changes are reviewed before non-engineers start making them. What it does right: a marketplace of providers and tools means the integration surface is not gated on the maintainers' time.

1. The knowledge pipeline handles ingestion, chunking and retrieval as its own concern rather than as nodes scattered through a graph, which keeps the retrieval configuration inspectable in one place. 2. Model management is separate again, so provider changes do not touch application logic. 3. Completed workflows are published as OpenAPI endpoints, which turns the visual artefact into a service any client can call.

That third property is the architecturally important one: the canvas is the authoring environment, not the runtime interface. Verification remains whatever node the author adds. No benchmark is published.

Annual pricing is an unusual choice here and a revealing one. Committing customers for a year rather than metering them monthly implies confidence in retention and a preference for predictable revenue over usage upside, which is how you build a company you can forecast rather than one you have to explain each quarter.

The distribution underneath it is exceptional and mostly unmonetised, which is the standard shape and the standard risk. Likely acquirer: a cloud provider or a data platform buying the application layer. Position: long, with the note that self-hosting is why most of those stars will never pay.

Two paths and the cheap one is not free. Running it ourselves means a database, a container platform and a named owner, which is roughly a quarter of an engineer indefinitely. The hosted Team plan at $1,590 a year is less than we spend on coffee, and identity federation and the controls I would need for sixty people are in the tier with no published price.

It integrates with nothing in our pipelines, so this is a product surface rather than a build tool. Approved with conditions: self-host, one owner, and no customer data in it until retention is settled.

Start with the licence, because the name is doing work. This is Apache-2.0 with extra conditions attached, which is not the same thing as Apache-2.0 and means the freedoms I assume by default need checking against the actual text before anything I build depends on them.

Everything else pleases me. One docker compose brings the whole platform up on my own hardware, the model layer accepts any OpenAI-compatible endpoint so my local server is a first-class provider, and it consumes MCP servers. Self-hosted, offline, with my own weights. The conditions in that licence are the only reason this is not a ten.