The Postgres-based alternative to Onyx
Put AI agents to work on company data with permission inheritance in the open-source core, standalone connector SDKs, and a leaner Postgres-based stack for self-hosted teams.
| Capability | Omni | Onyx |
|---|---|---|
| Permission-aware search and agents | Core OSS feature | Paid feature; not included in the free OSS edition |
| Workplace connectors | 20+ connectors; permissions are baked into OSS connector implementations | 40+ prebuilt connectors; source permission sync is paid, not OSS |
| Custom connectors and extensibility | Standalone SDK connectors in Python, TypeScript, or Rust; source is open | In-repo Python connector classes registered in backend and frontend code |
| Deployment model | Self-hosted today; hosted cloud coming soon | Self-hosted and hosted cloud available today |
| Operational complexity | Lower full-stack resource floor: 4 cores / 8 GB RAM | Lite is limited; Standard preferred: 8+ vCPU / 16+ GB RAM |
| AI agents and workflows | Agents with code execution; dedicated deep research mode coming soon | Deep research, code interpreter, knowledge graphs |
| Model and inference control | OpenAI-compatible (hosted + local) | Hosted providers, Ollama, LM Studio, LiteLLM Proxy |
| Search and retrieval architecture | Postgres-first retrieval; fewer moving parts | Standard mode uses separate search/index services |
| Admin and security controls | SSO in Business; SCIM and audit logs in Enterprise | Google OAuth in Business; OIDC/SAML and SCIM in Enterprise |
| Open source and source-code control | Apache 2.0 core | MIT core |
| Pricing and procurement | $0 self-host · $10/user/mo Business | $0 self-host · $20/user/mo Cloud, billed annually |
Permission-aware search and agents
Omni
Core OSS feature
Onyx
Workplace connectors
Omni
Onyx
Custom connectors and extensibility
Omni
Onyx
Deployment model
Omni
Onyx
Operational complexity
Omni
Onyx
AI agents and workflows
Omni
Agents with code execution; dedicated deep research mode coming soon
Onyx
Deep research, code interpreter, knowledge graphs
Model and inference control
Omni
Onyx
Search and retrieval architecture
Omni
Onyx
Admin and security controls
Omni
Onyx
Open source and source-code control
Omni
Onyx
Pricing and procurement
Omni
Onyx
Where Omni stands out
6 reasons teams choose Omni
Permission inheritance in the OSS core
AI agents are only useful on company data if they respect the same access model as the source system. Omni treats inherited permissions as a core open-source feature: permissions are baked into the OSS connector implementations, and connectors can emit document ACLs and group membership changes directly through the SDK protocol. In Onyx, comparable source permission sync is a paid feature, not part of the free OSS edition.
- Document permissions are part of the open connector event model
- Group membership sync is available to OSS connector authors
- Security review can inspect the permission pipeline directly
Postgres-based operations
Omni starts from Postgres and pgvector, which keeps the default deployment smaller and easier to reason about. The same database handles application data and retrieval for the common self-hosted path, so teams can monitor, back up, tune, and upgrade one familiar system before introducing specialized infrastructure. For teams without a dedicated platform group, fewer moving parts matter.
- Postgres-first architecture reduces infrastructure sprawl
- Lower CPU and RAM floor for small self-hosted deployments
- One familiar default database to operate, back up, monitor, and tune
Connectors built for extension
Omni supports 20+ connectors across Google Workspace, Microsoft 365, knowledge bases, collaboration tools, and project management systems. More importantly, the connector model is designed for teams to extend: standalone SDK connectors can be written in Python, TypeScript, or Rust, self-register by manifest, and expose sync, actions, OAuth, search operators, and MCP tools without merging code into the core app.
- Build connectors as standalone services instead of monolith patches
- Expose both indexed data and agent actions from the same connector contract
- Use MCP to turn existing tool servers into Omni connector capabilities
Licensing you can read in one sitting
Omni's Community edition is cleanly Apache 2.0 in one tree, and the Enterprise image is a single license away. If you are redistributing, embedding, or modifying the product, the boundary is plain: the OSS core is open, and commercial features ship in the licensed image. That keeps legal review straightforward for teams that want to build on top of the product.
- Community edition: one tree, Apache 2.0, no carve-outs
- Enterprise image: single commercial license, clear scope
- No ambiguity about what can be redistributed
Transparent self-hosted pricing
Omni's Business tier is $10/user/month billed annually, SSO included, self-hosted, and published publicly. The Community edition is free forever. Teams can start with OSS, upgrade when they need commercial features, and understand the cost before they talk to anyone. That makes pilots and security reviews easier to plan.
- Omni Business: $10/user/mo annual, SSO included, self-hosted
- Omni Community: $0, no seat minimum
- Enterprise controls can be added without changing the deployment model
Built to be extended around your workflows
Omni is open source software your team can adapt around internal tools, proprietary systems, and company-specific workflows. Build custom connectors, adjust retrieval behavior, inspect permission handling, and shape the product around the way your organization actually works. The goal is not just to search company data; it is to put AI agents to work on it.