Dify Statistics.

Dify’s public footprint in 2026 is easiest to misread if every large number gets treated as “users.” The open-source project is clearly large: the langgenius/dify repository had 147,742 GitHub stars, 23,264 forks, and about 1,375 contributors in the July 5, 2026 GitHub API snapshot. Those are strong open-source attention and participation signals.

The harder question is what those numbers prove. GitHub stars do not measure Dify Cloud users. Docker pulls do not equal active self-hosted deployments. A plugin repository is not the same as plugin adoption. Pricing pages do not reveal revenue. Funding is company momentum, not customer count. The useful way to read Dify in 2026 is as a layered platform: open-source project, self-hosted distribution, hosted workspace product, plugin and model-provider ecosystem, and LangGenius company context.

Where Dify Stands In 2026

The headline Dify numbers use different denominators, so read them as separate open-source, distribution, pricing, and company signals rather than one figure.

Stars Are Not Users: Reading Dify’s Numbers

Dify has a stronger public data trail than many private AI application-platform companies, but the trail is uneven. The open-source layer is visible through GitHub, release pages, contributor graphs, and Docker Hub. The hosted-product layer is visible through pricing, billing docs, model-provider docs, and workspace integrations. The company layer is visible through Dify’s funding announcement, Business Wire, and compliance announcements.

Those layers answer different questions. GitHub stars answer, “How much public developer attention does the project have?” Forks answer, “How many GitHub accounts copied the repository?” Contributors answer, “How broad is participation in the open-source project?” Docker pulls answer, “How often have official images been pulled?” Pricing answers, “How does Dify package the hosted product?” Funding answers, “What kind of investor and company momentum has been disclosed?”

None of those numbers answers, “How many active Dify Cloud users are there?” They also do not answer paid-customer count, monthly active workspaces, active self-hosted installations, revenue, retention, churn, net revenue retention, or customer deployment depth. That absence is not unusual for a private company with an open-source product, but it changes how each statistic should be used.

Open-Source Footprint: Stars, Forks, Contributors, Commits, And Releases

Dify’s open-source footprint is the strongest public part of its data story. The main langgenius/dify repository is not a small side project: the July 5, 2026 GitHub API snapshot showed 147,742 stars, 23,264 forks, 816 watchers/subscribers, and a latest push timestamp on the same day. The GitHub page rounded that to 148k stars and 23.3k forks.

The historical slope matters. Dify’s official milestone post says the project crossed 100,000 GitHub stars on June 5, 2025, and that Dify had been open-sourced on May 15, 2023. By July 5, 2026, the project had grown to about 147.7k stars in the API snapshot — roughly 47,700 stars above that official milestone.

Participation is also visible. The contributors endpoint pagination indicated roughly 1,375 contributors, and the commits endpoint indicated 11,474 commits, matching the web UI. These are signs of a busy open-source project, but they still measure project activity rather than production usage.

The issue and pull-request backlog needs careful interpretation. The repository API reported 883 open issues and PRs combined. A separated GitHub search returned 282 open issues and 601 open pull requests, plus 17,922 closed issues and 14,598 closed PRs. A large backlog can indicate active community engagement, heavy development, product complexity, or maintenance pressure — it should not be simplified into either “healthy” or “unhealthy” without looking at labels, age, and close rate.

Release state is easier to read. Dify 1.15.0 was the latest stable release in the July 5 snapshot, published on June 25, 2026, and the release page also shows 2026 releases such as 1.13.0, 1.14.0, and their patch versions. Dify 1.15.0 introduced difyctl, a command-line client for running Dify apps and workflows from terminal, scripts, and CI pipelines. That is a product-surface expansion from web console to developer automation.

Self-Hosting And Docker Distribution Signals

Dify’s self-hosting story is visible from two directions: official docs and official Docker images. The GitHub README says the easiest way to start the server is through Docker Compose, and the dedicated self-host docs show the quick start for Docker Compose. The docs tell users to clone the repository, enter the docker directory, copy the environment file, and run the compose command.

The same quick-start docs say the stack starts 5 core services and 6 dependent components: core services such as API, background job service, scheduler, web, and plugin daemon, plus components such as Weaviate, PostgreSQL, Redis, Nginx, SSRF proxy, and sandbox. That architecture matters for metrics — a single working environment can pull several images, and those images can be pulled again during upgrades, CI jobs, cache misses, failed deployments, and test runs.

The mistake is to sum those four counts and call the result deployments. If dify-api, dify-web, dify-sandbox, and dify-plugin-daemon are all part of the same Dify deployment pattern, then adding pull counts across services double-counts or quadruple-counts deployment-related activity. Docker pulls also include automated rebuilds, retries, cache warming, and version churn. Use them for distribution heat, not active instances.

The environment variable docs describe public console API URLs, backend API URLs, web app URLs, trigger URLs, file URLs, OAuth callback needs, collaboration mode, database settings, queue settings, background-job settings, and vector store profiles. These are real deployment surfaces that make Dify relevant for teams that want control over data, model keys, integrations, and infrastructure. The self-hosting signal is therefore strong but indirect: Docker Hub proves distribution activity around official images, and the docs prove Dify supports self-hosted operational deployment — neither proves how many active self-hosted installations exist today.

Dify Cloud, Pricing, And Workspace Packaging

Dify Cloud has a clearer public packaging surface than most open-source AI app platforms. The official pricing page lists Sandbox as free, Professional at $59 per workspace/month, Team at $159 per workspace/month, and Enterprise through contact and sales flows. Dify also says annual billing saves 17%, and notes that subscription prices exclude applicable taxes.

The billing docs make the unit explicit: billing is workspace-scoped. A subscription sets the team-member limit, feature access, and usage quotas for the entire workspace. That matters because Dify is not pricing simply by individual user seat — it is packaging a workspace with credits, apps, knowledge limits, storage, request limits, trigger events, workflow behavior, logs, and API access.

These plan limits are valuable for buyers because they show the shape of Dify Cloud’s product. They also show why pricing should not be used to infer revenue. A workspace can use different models, bring-your-own-key provider keys, Dify credits, different app volumes, and different workflow patterns. Discounting, annual billing, education pricing, enterprise contracts, and self-hosting all complicate any revenue inference. The public page gives enough detail to compare Dify’s hosted product with self-hosting, but it does not disclose active workspaces, paid customers, ARR, retention, or customer concentration.

Product Surface: Apps, Workflows, Agents, Knowledge, APIs, Logs, And CLI

Dify’s product surface is broader than a chatbot UI. The homepage says Dify offers agentic workflows, RAG pipelines, integrations, and observability in one place, and the GitHub README describes Dify as an open-source LLM app development platform with workflow, model support, Prompt IDE, RAG pipeline, agent capabilities, LLMOps, and backend-as-a-service APIs.

The docs make the app model more precise. Dify’s key-concepts page says the product is made for agentic app building, and that published apps can be accessed through API, web, or as an MCP server. It says the two main app types are Workflow and Chatflow, with Chatbot, Agent, and Text Generator as additional basic app types, and that those app types run on the same workflow engine underneath.

The Agent docs describe agents as chat-style apps where the model can reason through a task, decide what to do next, and use tools autonomously, and state that agents keep up to 500 messages or 2,000 tokens of conversation history before older messages are removed. Knowledge and RAG are also central: Dify’s knowledge docs describe quick create, knowledge pipelines, external knowledge bases, content management, retrieval testing, metadata, and app integration. The logs docs say conversation logs provide visibility into every interaction, helping teams debug issues and understand user behavior patterns.

The safest product conclusion is that Dify is an AI application platform, not only an agent demo or a RAG UI. It covers app creation, workflow orchestration, model selection, knowledge, APIs, logs, plugins, self-hosting, and CLI invocation. None of that proves usage by itself, but it explains why the open-source and Docker metrics are meaningful.

Plugin, Model Provider, And Integration Ecosystem

Dify’s ecosystem story changed materially with plugins. The official dify-official-plugins README says that starting with Dify v1.0.0 in February 2025, models and tools were migrated into plugins and stored in the official plugin repository, and those official plugins are uploaded to the Dify Marketplace and maintained by the official Dify team.

A July 5, 2026 GitHub tree snapshot of langgenius/dify-official-plugins found 260 top-level plugin-type directories under models, tools, datasources, extensions, triggers, and agent strategies. The official dify-official-plugins repository itself had 596 stars, 859 forks, and 44 open issues in the API snapshot. Dify’s plugin launch post frames the system as a mix of official plugins, partner solutions, and verified community contributions, and says Marketplace plugins undergo code review, run in isolation with defined permissions, and include data-handling declarations.

That 260 count should be used carefully. The recursive repository tree also had many nested folders, and those nested counts are not exact plugin counts — they are a map of implementation surface. The plugin tooling layer is smaller but active: the official dify-plugin-sdks repository had 143 stars, 140 forks, and 73 open issues; PyPI dify-plugin was at version 0.9.1 with 142 releases and a latest upload on June 17, 2026; and the npm package dify-client was at version 3.1.0.

This ecosystem surface matters because Dify is competing to be the orchestration layer between teams, models, data, tools, and apps. But ecosystem surface is not adoption. A plugin directory does not prove an active install. A package version does not prove downloads. A model-provider page does not prove model usage volume. Use these numbers to show platform breadth, then keep the usage gap visible.

Company Context: Funding, Valuation, And Enterprise Readiness

Dify’s company context became clearer in March 2026. Dify’s own blog says it raised a $30 million Series Pre-A led by HSG, with participation from GL Ventures, Alt-Alpha Capital, 5Y Capital, Mizuho Leaguer Investment, and NYX Ventures, and says returning investor 5Y Capital doubled down.

The company signal should stay in its lane. A $30M Series Pre-A is not revenue. A $180M valuation is not ARR. Investor participation is not customer count. The funding does show that Dify has investor-backed momentum around the enterprise agentic-workflow category, but it does not fill the missing denominators around active users, paid workspaces, self-hosted deployments, or retention.

Security and compliance add another enterprise-readiness layer. Dify says it completed SOC 2 Type II, ISO 27001:2022, and GDPR compliance for the second consecutive year, with audits covering personnel management, vendor onboarding, data security, system operations, and development security. Compliance is meaningful for buyers evaluating hosted or enterprise AI infrastructure, but it does not prove the number of enterprises using Dify. It proves that Dify is investing in the trust surface enterprises usually require before procurement can scale.

Market Context: AI App Builders And Agentic Workflows

Broader AI adoption explains why Dify’s category matters, but it should not be turned into Dify-specific adoption. The Stack Overflow 2025 Developer Survey says 84% of respondents use or plan to use AI tools in development, and that 51% of professional developers use AI tools daily. That is developer-AI category context, not Dify share.

GitHub’s Octoverse 2025 frames AI, agents, and typed languages as major shifts in software development. Dify’s open-source attention lives inside that broader GitHub ecosystem, but GitHub’s platform-level metrics do not prove Dify usage. The Stanford AI Index 2026 is useful for the same reason: it shows a broader AI adoption and measurement environment, and its framing around rapid generative AI adoption supports cautious language around agentic platforms. McKinsey’s State of AI notes that agent scaling is still uneven by function.

The key category takeaway is that Dify sits in a market where AI app building is moving from prompts and chat interfaces toward workflows, agents, model routing, knowledge retrieval, tools, APIs, and observability. Dify’s own docs line up with that shift through workflows and chatflows, agents, knowledge, integrations, and API publishing.

What This Means For Operators And AI Builders

For operators, the most important Dify statistic is not one number. It is the gap between visible open-source momentum and invisible production usage.

For GTM teams, Dify’s company context should be used carefully. The $30M Series Pre-A, $180M valuation, and SOC 2 / ISO 27001 / GDPR posture support the idea that Dify is serious about enterprise workflows. They do not reveal conversion, average contract value, or churn.

The strongest operator takeaway is that Dify has public evidence of momentum at several layers: open-source attention, release velocity, Docker distribution, hosted pricing, plugin architecture, security and compliance posture, and venture backing. The missing public data is equally important: exact active users, paid customers, revenue, retained cohorts, active self-hosted deployments, and customer distribution are not available in a clean current public dataset.

What To Watch Through The Rest Of 2026

The first number to watch is GitHub velocity. If stars, forks, contributors, releases, and closed PRs keep rising, the open-source project remains active; if releases slow or the backlog grows without resolution, the same public metrics could start to mean maintenance pressure rather than momentum. The second is Docker distribution by image — pull counts should remain pull counts, but changes in update cadence and tags can indicate active packaging.

The current 260 top-level plugin-type directories show breadth in source control, but marketplace install and usage data would be a much stronger adoption signal. Dify has disclosed $30M in Series Pre-A funding and Business Wire reports a $180M valuation. A future disclosure of revenue, paid workspaces, enterprise customers, retention, or active deployments would materially change the data picture.

Frequently Asked Questions

Sources and Further Reading