AxonHub is an open-source AI gateway, written in Go and released under the Apache-2.0 licence, that lets an application call 100+ large language models through whichever AI SDK it already uses and switch providers without changing code — built for developers and platform teams who need one self-hostable control point for model access, failover, cost and tracing.

What it is

AxonHub sits between an application and the model providers, in the "Developer Tools / API Development & Testing" category. It accepts requests in OpenAI, Anthropic or Gemini API formats and translates them transparently into whatever format each supported provider expects, so an OpenAI SDK call can reach Claude and an Anthropic SDK call can reach GPT. The project describes itself as an all-in-one AI development platform whose core promise is "any SDK, any model, zero code changes" — switching providers is a configuration change rather than a refactor.

The concrete problem it replaces is the glue code and vendor lock-in that accumulate when an application talks to several model providers at once. The README names four failures it targets: provider lock-in (moving from GPT-4 to Claude or Gemini instantly), integration complexity (one API format against 10+ providers), observability gaps (complete request tracing out of the box), and cost control (real-time usage tracking and budget management). Rather than maintaining a separate client, key rotation scheme and logging path per vendor, a team points existing SDKs at AxonHub and manages providers centrally.

Key capabilities

• Request translation between OpenAI, Anthropic and Gemini API formats, documented under docs/zh/api-reference/openai-api.md, anthropic-api.md and gemini-api.md.
• Text generation, image generation, rerank and embedding APIs are marked done and documented as image-generation.md, rerank-api.md and embedding-api.md; the realtime API is still listed as Todo.
• Built-in failover and intelligent load balancing, covered by docs/zh/guides/load-balance.md.
• Complete request tracing with thread-level observability, so a full request timeline can be inspected to locate problems faster (docs/zh/guides/tracing.md).
• Enterprise-grade RBAC with fine-grained access control, usage quotas and data isolation (docs/zh/guides/permissions.md).
• Cost control through real-time usage tracking and budget management.
• A dashboard covering channels, model pricing, models, trace viewing and request monitoring.
• Supported providers include OpenAI, Anthropic, Zhipu (GLM-4.5), Moonshot (kimi-k2), DeepSeek (DeepSeek-V3.1), ByteDance Doubao, Gemini, Jina AI, OpenRouter and ZAI, with AWS Bedrock in testing.

Who uses it and how

• Teams running AI coding agents and CLI tools such as Claude Code and Codex, which the topic list calls out, routing them through one gateway instead of per-tool provider configuration.
• Platform teams that need per-department access control, usage quotas and data isolation delivered by the RBAC layer.
• Engineering organisations that want end-to-end tracing and request monitoring across all model traffic rather than per-provider logs.
• Cost-conscious operators who need real-time usage tracking and budget management over many providers and keys.
• Groups that want to evaluate AxonHub without local setup, running it in a browser or on a phone through Bloome and sharing that instance with a team.

Getting started

The project advertises Docker readiness, and a hosted instance is available at https://axonhub.onrender.com/ alongside the Bloome option for launching it from a browser or phone without configuring a local environment.

How it compares

No list of paid products that AxonHub replaces is given in the facts, and no directly comparable open-source AI gateway is named there either, so it stands alone in this registry. The only third-party services described are sponsors — API relay and MaaS offerings such as APIKEY.FUN, Infistar.cc, Qiniu Cloud AI and FennoAI — which are commercial upstream providers rather than peer projects. Readers looking for a head-to-head comparison will not find one in the material supplied here.

When to use it — and when not to

Choosing AxonHub means operating a Go gateway yourself and keeping it between your application and every provider, which is real infrastructure to run; the facts do not spell out the backing database, object storage or SMTP requirements, so those should be confirmed from the repository before committing. The project is Apache-2.0 licensed and actively pushed, but the README is chiefly Chinese with English and Japanese translations and documentation links pointing at the docs/zh tree, the realtime API is still Todo, AWS Bedrock is only in testing, and 112 open issues suggest moving parts. Teams that need only a single provider, or who cannot accept an additional hop in the request path, are better served by calling that provider directly.

    APIKEY.FUN 是一家专业的企业级 AI 中转站，致力于为企业和个人开发者提供稳定、高效、低成本的 AI 模型 API 接入服务。<br/>
    平台支持 Claude、OpenAI、Gemini 等主流热门模型，价格低至官方原价的 7%。<br/>
    AxonHub 用户通过<a href="https://apikey.fun/register?aff=axonhub" target="_blank">专属链接注册</a>，还可享受最高充值永久 95 折专属优惠。

</td>

    ⚡ 企业级高并发稳定调度｜官方 1 折起 · 透明倍率 · 多节点冗余<br/>
    🧠 全系模型支持｜ChatGPT · Claude · Gemini · DeepSeek<br/>
    🎨 全模态接入｜文本 · 图像 · 视频 · Embedding / Rerank


    AxonHub 用户通过<a href="https://lj.s.gy/DUq59H" target="_blank">专属链接</a>注册，可享专属特惠。

</td>