Build AI assistants and deploy autonomous agents — all from one self-hosted platform. Define multi-step workflows in plain YAML, connect 2,673+ enterprise tools, and run agents anywhere with a single binary. No LangChain. No LangGraph. No code.
Open Enterprise ships as a single Docker container with all capabilities included. No modules, no add-ons, no per-seat pricing surprises.
Isolated workspaces with their own knowledge bases, system prompts, LLM settings, and user access controls. Teams stay separated, data stays clean.
Build and deploy autonomous agents with tool calling, cron scheduling, and multi-agent chains. Agents can read connectors, process documents, and trigger actions.
The largest open-source connector catalog — databases, CRMs, DevOps, healthcare, legal, finance, and more across 45+ categories. Sync data automatically into your workspace knowledge base.
Upload documents, sync connectors, and let your AI cite sources from your own data. Supports PDFs, DOCX, Markdown, HTML, CSV, and 24+ other formats.
Built-in data loss prevention policies. Block, redact, or audit sensitive content — passwords, API keys, credit card numbers, and custom patterns.
Super admin dashboard with user management, workspace oversight, activity logs, token usage tracking, and instance-wide settings.
Choose from 8 vector DB providers: LanceDB (built-in), Pinecone, Weaviate, Qdrant, Chroma, Milvus, Zilliz, and Astra DB.
Embed a chat widget on any website with a single script tag. White-label it for your product, docs site, or internal portal.
Full REST API with key-based authentication. Integrate Open Enterprise into your existing workflows, CI pipelines, or custom applications.
Most AI agent frameworks require you to write Python, wire up LangGraph nodes, or manage complex orchestration code. Open Enterprise is different — define any agent workflow in a simple YAML file. Steps, connectors, schedules, and logic are all declared, not coded.
No framework to learn. No Python environment. Just a YAML file and your LLM API key.
The instructions field is the agent's system prompt. Describe what it does and how it should behave.
List which tools the agent can use. Credentials live separately in oe-config.json — your YAML never contains secrets.
Break complex workflows into named steps. Use {{param}} to pass values at runtime without editing the file.
Execute with OE Runtime for local use, or deploy to the Platform for scheduling, RBAC, and multi-user access.
Open Enterprise is fully open source. Audit the code, fork it, extend it, or contribute back. Your data never leaves your servers unless you choose to send it to an external LLM.
Zero phone-home. No usage data sent to us. Run it air-gapped if you need to.
Connect OpenAI, Anthropic, Azure, Ollama, LM Studio, Together, or 11 other providers. Switch anytime.
The core platform is freely available on GitHub. Clone it, deploy it, ship it.
Permissive for internal use. Modifications that you distribute must stay open.
Read every line of code handling your data. No black boxes, no hidden behaviour. Security teams love it.
Ships as a single ~200 MB Docker image. Runs on a $6/month VPS. No Kubernetes required unless you want it.
Built in the open with community contributions. File issues, submit PRs, shape the roadmap directly on GitHub.
2,673 connectors across 45+ categories — the largest open-source connector catalog. Sync data from your entire stack into your AI knowledge base automatically.
From zero to a fully operational AI platform with a single Docker command.
Pull openenterprise/open-enterprise-community from Docker Hub. No build step required.
Set JWT_SECRET, SUPER_ADMIN_EMAIL, and SUPER_ADMIN_PASSWORD. LLM provider and API keys are configured from the admin panel after first login.
Log in as super admin, create workspaces, invite users, and connect your data sources.
Build agents with your connectors and schedule them. AI starts working for your team automatically.
| Frontend | React + Vite |
| Backend | Node.js + Express |
| Database | SQLite / PostgreSQL |
| File Storage | Local filesystem |
| Queue | In-memory |
| Vector DB | LanceDB (default) + 7 others |
| LLM Providers | 17+ (OpenAI, Ollama, Anthropic, etc.) |
| Embedding | Any provider (OpenAI, Ollama, etc.) |
| Deployment | Single Docker container |
| Process Manager | PM2 |
From foundation to ecosystem — a complete agentic runtime, self-hosted and open source.
Multiple LLMs, Embeddings, Vector Databases
RAG, Memory, Context Engineering
AI Assistant / Chat, Workspaces (Multi-tenant)
Tools / Connectors, Tool Calling, MCP
Chains, Conditional Branches, Multi-step Pipelines, Loops, Routers, State Management
Agents, Agent Planning, Multi-agent Collaboration, Reflection / Self-Critique
Automation, Event-driven Workflows, Human Approval, Runtime / Execution Engine
Guardrails / DLP
Observability (Token Usage, Activity Log, Dashboard)
Marketplace
OE Runtime is a standalone binary that executes any Open Enterprise agent YAML locally — on Windows, Linux, or macOS. No Docker, no database, no UI. Just point it at a YAML file and run.
--param
Both run the same agent YAML files and use the same execution engine. The difference is what surrounds it.
| ⚡ OE Runtime | 🖥️ Platform | |
|---|---|---|
| Best for | Developers, local scripts, CI/CD | Teams, orgs, production |
| Install | Single binary — no install | Docker or source |
| UI | CLI only | Full web application |
| Multi-user + RBAC | — | ✓ Workspaces + roles |
| RAG | Roadmap | ✓ Full pipeline |
| Scheduling | External cron | ✓ Built-in cron |
| Human approval gates | Roadmap | ✓ Built-in |
| Platforms | Windows · Linux · macOS | Any Docker host |
Both use the same agent YAML format — author and test locally with Runtime, deploy to the Platform for your team.
Most AI agent frameworks require you to write Python. We don't. Here's the architectural difference.
| Open Enterprise | LangGraph | CrewAI | AutoGen | Dify | |
|---|---|---|---|---|---|
| Agent definition | YAML | Python | Python | Python | UI + YAML |
| No-code | ✓ | ✗ | ✗ | ✗ | Partial |
| Standalone binary | ✓ | ✗ | ✗ | ✗ | ✗ |
| Built-in connectors | 2,673 | 0 (write your own) | 0 (write your own) | 0 (write your own) | ~60 |
| Built-in RAG | ✓ | ✗ | ✗ | ✗ | ✓ |
| Human approval gates | ✓ | Code required | Code required | Code required | ✓ |
| MCP support | ✓ | ✓ | ✗ | ✗ | Partial |
| Agent marketplace | ✓ | ✗ | ✗ | ✗ | ✓ |
| License | Apache-2.0 | MIT | MIT | MIT | Apache-2.0 |
Architectural comparison only — every tool excels in different contexts.
The Community Edition is free forever — but when your organization needs SSO, DLP, white-label branding, multi-tenant workspaces, and dedicated support, Open Enterprise Commercial is built for you.
Share agent YAMLs, get help with connectors, follow the roadmap, and talk to the team. Whether you're just getting started or building in production — you belong here.
Self-host for free, or let us manage your instance so you can focus on building.