Repolode
Submit
ogx-ai

ogx-ai/ogx

Open-source agentic API server that exposes OpenAI-compatible endpoints and routes requests to any model or backend.

Open-source alternative to:OpenAI API

View on GitHubWebsite
Stars
8,444
Forks
1,394
Open issues
141
Last commit
1 hour ago

What is it?

OGX is an open-source API server for building AI applications. It acts as a drop-in replacement for the OpenAI API, so existing OpenAI-compatible clients and agent frameworks can run against it without code changes. The server can be run on a laptop, in a datacenter, or in the cloud.

The project uses a pluggable provider architecture. Developers can use Ollama locally and deploy the same application against vLLM or a managed model service, with the API surface staying the same. It also provides a Responses API with server-side tool calling, MCP server integration and built-in file search, along with vector stores, files, and batch processing endpoints.

OGX additionally serves Anthropic SDK (/v1/messages) and Google GenAI SDK requests natively, and ships official Python and TypeScript client SDKs. It was formerly known as Llama Stack.

Key features

  • 01Drop-in OpenAI-compatible endpoints for chat completions, completions and embeddings.
  • 02Responses API with tool calling, MCP server integration and built-in file search.
  • 03Vector stores, files and batch processing endpoints.
  • 04Native Anthropic and Google GenAI SDK endpoints alongside the OpenAI API.
  • 05Pluggable providers so the same API can run against Ollama, vLLM or managed services.

Use cases

  • Running an OpenAI-compatible backend locally while developing, then deploying the same code to production.
  • Switching between open and proprietary models without rewriting application code.
  • Building agents that call tools and MCP servers through a single Responses API call.
  • Hosting a self-managed AI gateway inside your own infrastructure.

Quick start

  1. 1

    Install the OGX package with uv

    uv pip install ogx
  2. 2

    Start the OGX server (auto-detects providers from the environment)

    uv run ogx go

Things to know

  • Needs at least one model provider or backend configured to serve real inference.

In collections

Open-source AI agent frameworks & appsMCP serversSelf-hosted apps

Topics

#openai-compatible#llm#agents#api-server#rag#vector-database#mcp#self-hosted

FAQ

What is ogx?
Open-source agentic API server that exposes OpenAI-compatible endpoints and routes requests to any model or backend. OGX is an open-source API server for building AI applications. It acts as a drop-in replacement for the OpenAI API, so existing OpenAI-compatible clients and agent frameworks can run against it without code changes. The server can be run on a laptop, in a datacenter, or in the cloud.
Is ogx free to use? What license does it have?
ogx is open source under the MIT license. Check the license terms before using it in commercial products.
Is ogx still maintained?
Actively maintained. The last commit to ogx was on Oct 22, 2026.
What can I use ogx for?
Running an OpenAI-compatible backend locally while developing, then deploying the same code to production. Switching between open and proprietary models without rewriting application code. Building agents that call tools and MCP servers through a single Responses API call. Hosting a self-managed AI gateway inside your own infrastructure.
How do I install ogx?
Install the OGX package with uv: uv pip install ogx Start the OGX server (auto-detects providers from the environment): uv run ogx go
Where can I get ogx?
The source code and installation instructions are on GitHub: https://github.com/ogx-ai/ogx

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects