ogx-ai
OGX (Open GenAI Stack) is an open-source AI application server and Python library for building production generative-AI and agentic applications. It provides a unified, multi-SDK API layer over inference, vector stores, file storage, moderation, tool calling, and server-side agentic orchestration, with pluggable providers and deployment anywhere.
At a Glance
- AI application developers and engineering teams
- Platform and infrastructure teams
- Enterprise teams requiring multi-tenancy, access control, observability, and on-premises deployment
- Developers using local/self-hosted model infrastructure
- +1 more
AI Tools by ogx-ai
(1)
OGX
Open Source AI Application Server
Discussions
No discussions yet
Be the first to start a discussion about ogx-ai
Latest News
Multi-Tenant AI Infrastructure with OGX: Tenant Isolation, ABAC, and Defense in Depth
The OGX Python SDK Has a New Foundation
Connect Codex CLI to Local and Hosted Models with OGX
Under the Hood: How OGX Enforces Guardrails Inside the Agentic Loop
Products & Services
HTTP server that exposes OpenAI-compatible endpoints plus native Anthropic Messages and Google Interactions APIs; it routes to pluggable providers and runs server-side agentic loops including RAG, MCP tool use, multi-step reasoning, conversation state, moderation, and observability. Source: https://ogx-ai.github.io/blog/ogx-v1
In-process/library mode with the same core capabilities as the server, intended for scripts, notebooks, and latency-sensitive applications. Source: https://ogx-ai.github.io/
Python client SDK for interacting with OGX. Source: https://github.com/ogx-ai/ogx-client-python
TypeScript client library for OGX. Source: https://github.com/ogx-ai/ogx-client-typescript
Market Position
OGX positions itself as more than an inference gateway: it is a server-side agentic application server combining inference routing, API translation, RAG, MCP, storage, moderation, and orchestration. Its differentiators are native OpenAI, Anthropic, and Google SDK/API compatibility, a multi-provider architecture, and server-side tool/RAG loops; it explicitly distinguishes itself from framework-style libraries such as LangChain or LlamaIndex and from simple proxy/gateway products. Source: https://ogx-ai.github.io/blog/from-llama-stack-to-ogx
Founding Story
OGX grew out of Meta's Llama Stack project, which began as an API-standardization effort around Llama models. The project was renamed and moved to the ogx-ai GitHub organization because its scope had expanded to 23 inference providers, multiple SDK/API surfaces, and a server-side agentic loop; the team wanted a neutral name that did not imply Llama-only support or a framework. The rename announcement was signed by Charlie, Francisco, Matt, Raghu, and Seb. Source: https://ogx-ai.github.io/blog/from-llama-stack-to-ogx
Business Model
Revenue Model
OGX is distributed as open-source software under the MIT license and is self-hosted or embedded by users; the public sources describe no hosted service, subscription, usage billing, or commercial pricing. Source: https://github.com/ogx-ai/ogx/blob/main/LICENSE
Target Markets
- AI application developers and engineering teams
- Platform and infrastructure teams
- Enterprise teams requiring multi-tenancy, access control, observability, and on-premises deployment
- Developers using local/self-hosted model infrastructure
- Organizations using OpenAI, Anthropic, Google, AWS, Azure, NVIDIA, or other model providers
- Building production generative-AI applications
- Agentic applications requiring server-side planning, tool execution, and synthesis
- Retrieval-augmented generation over uploaded documents
- Connecting coding assistants such as Claude Code, Codex CLI, and OpenCode to local or hosted models
- Running one API surface across local models, self-hosted inference, commercial APIs, and cloud platforms
- Multi-tenant AI platforms serving multiple teams or customers