# AxonHub

> Open-source AI gateway that lets you use any SDK (OpenAI, Anthropic, Gemini) to call 100+ LLMs with built-in failover, load balancing, cost tracking, and end-to-end request tracing.

AxonHub is an open-source AI gateway built in Go that sits between your application and AI model providers, transparently routing requests without requiring any code changes. It is licensed under Apache-2.0 (with LGPL-3.0 for the LLM transformer module) and is actively developed by the looplj team on GitHub. A live demo instance is available at axonhub.onrender.com.

## What It Is

AxonHub is a unified API gateway for large language models. It accepts requests formatted for OpenAI, Anthropic, or Gemini SDKs and translates them on the fly to whichever backend provider you have configured — so you can call Claude using the OpenAI SDK, or call GPT-4 using the Anthropic SDK, with zero code changes. The core problem it solves is vendor lock-in: switching providers becomes a configuration change rather than a refactor.

## Supported Providers and API Types

AxonHub supports a broad and growing set of providers and modalities:

- **Text generation**: OpenAI (GPT-4, GPT-4o, GPT-5), Anthropic (Claude 3.5, Claude 3.0), DeepSeek (DeepSeek-V3.1), Gemini (Gemini 2.5), Moonshot (kimi-k2), Zhipu AI (GLM-4.5), ByteDance Doubao, OpenRouter, NanoGPT
- **Image generation**: OpenAI, Gemini, ByteDance Doubao, OpenRouter, ZAI, NanoGPT
- **Embeddings**: Jina AI, plus OpenAI-compatible embedding endpoints
- **Reranking**: Jina AI reranker
- **AWS Bedrock and Google Cloud (Claude on GCP)**: listed as in testing
- **Realtime conversation**: listed as planned (Todo)

## Architecture and Core Features

The gateway is built around a flexible transformer pipeline that normalizes request and response formats across providers. Key capabilities include:

- **Any SDK → Any model**: Use OpenAI, Anthropic, or Gemini SDK syntax interchangeably against any configured backend
- **Intelligent load balancing**: Sub-100ms automatic failover, always routing to the healthiest available channel
- **Enterprise RBAC**: Fine-grained access control, usage quotas, and data isolation per user or team
- **Full request tracing**: Thread-level observability with complete request timelines for faster debugging
- **Real-time cost tracking**: Per-request cost breakdown covering input tokens, output tokens, and cached tokens
- **Channel management**: Add multiple provider API keys, test connections, and enable/disable channels from a dashboard

## Deployment Model

AxonHub is self-hosted. The README provides a deployment guide and references a `deploy-axonhub` skill for agent-assisted deployment. Once running, users initialize the system via a setup wizard, add provider channels with their API keys, create AxonHub API keys for clients, and point their existing SDK `base_url` to the local AxonHub instance (e.g., `http://localhost:8090/v1`). Docker support is included. The project uses Go (gin framework), GraphQL (gqlgen), and an ORM (ent), with TiDB Cloud used for the demo deployment.

## Update: v1.0.0-beta10

The latest release is **v1.0.0-beta10**, published in September 2026. The repository was created in September 2025 and has seen continuous development, with the last push in September 2026. The project is tagged as beta, indicating active pre-1.0 development. The GitHub repository has accumulated over 5,000 stars and 700 forks according to the repository metadata, and appears on Trendshift as a trending repository. The default branch is named `unstable`, reflecting the active development cadence.

## Features
- Unified OpenAI/Anthropic/Gemini compatible API
- Transparent SDK-to-model translation with zero code changes
- Intelligent load balancing with sub-100ms failover
- Enterprise RBAC with fine-grained access control and usage quotas
- Full request tracing with thread-level observability
- Real-time cost tracking per request (input, output, cached tokens)
- Channel management for multiple AI providers
- Text generation, image generation, embeddings, and reranking support
- Docker-ready self-hosted deployment
- Support for 10+ providers including OpenAI, Anthropic, Gemini, DeepSeek, OpenRouter

## Integrations
OpenAI, Anthropic, Google Gemini, DeepSeek, Moonshot (Kimi), Zhipu AI (GLM), ByteDance Doubao, OpenRouter, Jina AI, AWS Bedrock, Google Cloud (Vertex AI), NanoGPT, ZAI

## Platforms
WEB, API, CLI

## Pricing
Open Source

## Version
v1.0.0-beta10

## Links
- Website: https://axonhub.onrender.com/
- Documentation: https://deepwiki.com/looplj/axonhub
- Repository: https://github.com/looplj/axonhub
- EveryDev.ai: https://www.everydev.ai/tools/axonhub