AI Potluck
Product / UX / Orchestration & agents

Dify

LangGenius

Platform for building LLM applications, combining a visual workflow builder, RAG pipelines, agent tooling and model management across providers in one workspace. It deploys to the vendor's cloud, a customer VPC or self-hosted infrastructure with a single Docker command.

The license is a custom derivative that GitHub declines to classify, which the openness axis resolves. Verified 2026-08-13 via the langgenius/dify repository and the Dify product site.

Openness

2 high confidence
2.0
license
Dify-Open-Source-License(Apache-2.0-based,non-OSI,added-conditions:multi-tenant/branding-restrictions)
source
public

The Dify Open Source License is Apache-2.0 with added conditions: a multi-tenant service restriction and branding requirements. That puts it outside OSI approval and charges for a class of use, so the published source does not make the product freely runnable - you can read it but not run it freely, which is what source-available scores here. Cloud and enterprise tiers sit alongside it. The vendor says as much in its own words, describing the product as source-available under an "Apache-2.0-derivative license", self-hostable with a single Docker command, next to its cloud tier; the repo publishes the whole runtime, and GitHub declines to place the license under a recognized SPDX badge.

  • https://github.com/langgenius/dify recorded 2026-08-13

    full runtime published; GitHub reports no SPDX license badge for the Dify Open Source License; 152.3k stars, 24.0k forks

  • https://dify.ai/ recorded 2026-08-13

    'source-available. Self-host with Docker. 152K+ stars on GitHub. Apache-2.0-derivative license', single-command Docker deploy, full agent runtime plus workflow

Adoption

3 high confidence
3.0

152,331 stars on langgenius/dify, alongside the Dify Cloud managed tier, which bands at >10K stars, level 3 on the stars scale. Dify ships as a Docker Compose stack with a managed cloud beside it, publishes no install, download or user figure, and has no first-party registry package, so stars are the only instrument available - and because a star is not a download, a star-based reading is capped at level 3.

Capability

4 high confidence
4.0

Production agentic-workflow builder: visual workflow, RAG pipeline, agent capabilities, model management across hundreds of LLMs, and built-in observability (Langfuse, Opik, Arize Phoenix). Rated on feature breadth. No benchmark placement exists for a workflow builder, so this stays a feature-matrix judgment, one below the openhands anchor.

  • https://github.com/langgenius/dify recorded 2026-08-13

    visual workflow builder, RAG pipeline, agent capabilities and model management published in full; no benchmark figure

  • https://dify.ai/ recorded 2026-08-13

    full agent runtime plus workflow, single-command Docker deploy, model management across providers

Verified 2026-08-13