Product

Dify

Open-source, low-code LLM app platform — visual agentic workflows, RAG and tool orchestration, 150k+ GitHub stars [1][2].

1. Core Product / Service

Dify (dify.ai, by LangGenius Inc.) is an open-source, low-code platform for building, deploying and operating LLM applications and agentic workflows [2]. It provides a visual flow builder (drag-and-drop nodes for RAG pipelines, agents, and tools), a prompt-engineering IDE, model management that connects to any LLM (OpenAI, Anthropic, open-weight models, or self-hosted), 1,000+ tool integrations, observability/APM over every call, and Backend-as-a-Service (BaaS) APIs that expose built apps to external code [3].

The project is one of the largest open-source AI repos on GitHub — the langgenius/dify repository passed 150,000 stars in mid-2026 (≈154,936 stars by Sept 2026, ranked #1 among RAG projects, 800+ contributors) [1][3]. It ships as a self-hostable Community Edition plus a managed Dify Cloud; the open-source version runs on 1.4M+ machines with 2,000+ teams [1].

2. Target Users & Pain Points

  • Application/agent developers who want to ship LLM features without hand-rolling orchestration, retrieval, and tool plumbing.
  • Enterprises running self-hosted, data-sovereign AI apps — 280+ paying customers including Volvo, ThermoFisher Scientific, Novartis, Maersk, and ETS [1][2].
  • Organizational multi-user teams (see onyx) evaluating agent platforms for shared, permissioned knowledge + workflow scenarios — the context of Jimmy's 2026-09-19 research comparing Dify against Onyx for Lark-based multi-tenant agent setups [local].

Pain solved: the gap between a raw model API and a production app — retrieval, state, evaluation, observability, and deployment are bundled so teams stop rebuilding the same orchestration layer.

3. Competitive Landscape

Platform Type Differentiation vs. Dify
Dify Open-source LLM app/agent platform Largest OSS community, visual workflow builder, BaaS APIs
onyx Open-source enterprise search + RAG assistant Knowledge/search-first (connectors, access control), not a general app builder
LangChain / LangGraph Open-source framework (code-first) Lower-level, developer-code, no managed UI/observability
Flowise Open-source visual flow tool Simpler, chat-flow only; smaller ecosystem
composio Agent tool-integration middleware Tool/auth layer, not a full app builder

Dify's edge is breadth and community gravity: it covers the whole app lifecycle (build → observe → deploy) where most competitors are either frameworks (code) or point tools [1][3].

4. Unique Observations

  • Open-source-first go-to-market. Dify was open-sourced (≈99%) from day one, letting a Chinese-origin product accumulate 150k+ GitHub stars and win global enterprise logos (Novartis, Volvo) before spending on a US sales force [1]. The community is the funnel.
  • "AI Infra" positioning + profitability. Founder Luyu Zhang frames Dify's edge as organizational — "we deliberately hire people big tech doesn't want" — and the company is reportedly already profitable at ~100 employees, unusual for a heavily-funded open-source vendor [1][3].
  • Silicon Valley relocation signals the global ambition. The company is moving HQ from China to Menlo Park, a bet that Western enterprise buyers will standardize on its platform the way they did on open-source infra [1].

5. Financials / Funding

  • Founded: 2023, by Luyu Zhang (张路宇) and John Wang [1].
  • Total raised: ~$41.5M (PitchBook); ~$30M commonly cited for the latest round [3].
  • Latest round: $30M at a ~$180M valuation (announced early 2026, styled "Series Pre-A"), led by HSG (formerly Sequoia Capital China), with GL Ventures, Alt-Alpha Capital, 5Y Capital, Hillhouse, and Mizuho Leaguer Investment [1][2].
  • Earlier rounds: $1.5M seed (Jul 2023), $6M (Aug 2024) [3].
  • Scale: ~100 employees; profitable; 280+ enterprise customers [1][2].

6. People & Relationships

  • Founder / CEO: Luyu Zhang (张路宇) — middle-school dropout, ex-Tencent (ran Tencent Cloud's DevOps "Coding" team, joined 2018); many early Dify hires came from that same team [1].
  • Co-founder: John Wang [1].
  • Competitors / peers: onyx, LangChain/LangGraph, Flowise, composio.

Related

Last compiled: 2026-09-21