Skip to content

opencodex Review

opencodex is a free, MIT-licensed local proxy that lets OpenAI Codex, Claude Code, Claude Desktop and Grok Build run on models from 40+ providers, with a web dashboard and desktop app.

Freemium · Web · API

Stork's verdict on opencodex

opencodex is a free, fast-moving way to run Codex and Claude Code on almost any model, best for developers comfortable configuring a local proxy.

opencodex reviewed by Stork AI · stork.ai/en/opencodex

Screenshots

720p. 480p is half that. Minimum 5 cents a video.
Ten minutes of talking video costs $2.
We run open models on our own GPUs. When our cost per minute falls, so does the price.
Make a photo talkDevelopers: there is an API

What opencodex does

opencodex is an open-source gateway that sits between your AI coding tools and the models behind them. It runs a small local HTTP server (default localhost:10100) that serves both the OpenAI Responses API that Codex speaks and the Anthropic Messages API that Claude Code speaks, then translates each request to whichever provider you route it to. The project says this works with zero client patches and keeps streaming, tool calls, reasoning and images where the target model supports them. Five adapters (Anthropic Messages, Google Gemini, Azure, an OpenAI Responses passthrough and any OpenAI-compatible Chat Completions endpoint) cover 40+ built-in providers, including Anthropic, Google, xAI, Kimi, Ollama Cloud, Groq, OpenRouter, DeepSeek and GLM, plus local models.

Setup is npm install -g @bitkyc08/opencodex, then ocx init to add a provider and wire it into your Codex config, ocx start to run the proxy and ocx claude to launch Claude Code through it; ocx stop restores native Codex. A beta desktop app bundles the proxy, a web dashboard and a tray icon for macOS 13+ (universal), Windows x64 (the installer is not code-signed yet, per the download page) and Linux (AppImage and .deb). Natively configured clients are Codex (CLI, app and SDK), Claude Code, Claude Desktop and Grok Build; other tools such as Qoder and WorkBuddy get an exported config, and anything that accepts a custom OpenAI or Anthropic base URL can point at the local port.

opencodex pricing

opencodex itself is free. The code is published on GitHub under the MIT License and installed from npm; there is no paid plan, account or hosted tier on the project site.

ItemPriceNotes
opencodex CLI and proxy (npm)$0MIT License
OpenCodex Desktop (beta)$0macOS, Windows, Linux downloads on the homepage
Model usageBilled by each providerYour own API keys, or logins such as ChatGPT, Anthropic, xAI or Kimi accounts

What you actually spend depends on the providers you route to. The docs flag a few places where opencodex can trigger paid calls: the Image Bridge, which sends image requests to xAI Grok Imagine, is off by default "to avoid unexpected xAI charges", and combos only fall back to a pay-per-use target marked metered when you explicitly allow it.

Prices checked on the vendor's pricing page on November 2, 2026. There is no separate pricing page; see the opencodex homepage and the MIT license on GitHub.

Key features

What the opencodex docs list:

  • Translates Codex (Responses API) and Claude Code (Messages API) requests to 40+ providers
  • OAuth login for xAI, Anthropic and Kimi accounts, ChatGPT login forwarding, or API keys
  • ChatGPT account pool: new Codex sessions route to the lowest-usage healthy account
  • Routed models appear in the Codex app model picker and Claude Code's /model picker
  • Combos: one virtual model with failover, round-robin, random or least-used strategies
  • Web search and vision sidecars for models that lack them; opt-in Grok Imagine image bridge
  • Local web dashboard for providers, accounts, models, sidecars and request logs
  • Docs in English, French, Korean, Simplified and Traditional Chinese, Russian, Japanese, Turkish

Who opencodex is best for

opencodex suits developers who like the Codex or Claude Code workflow but want to run it on other models: a cheaper model for routine tasks, a local model through Ollama, or a failover chain across several subscriptions. It is a local, single-user tool first; it can be exposed as a remote hub, but that needs explicit configuration and admin tokens. The project ships very often (v2.96.0 on November 1, 2026) and its docs are dense, so expect to read configuration notes. Before routing a consumer subscription login through any third-party proxy or pooling several accounts, check that provider's terms.

If you want a general-purpose LLM gateway for an application rather than for coding agents, look at LiteLLM or Portkey. If you only want one API key that reaches many models, OpenRouter is simpler. To run models entirely on your own machine, pair any of these with Ollama.

What users say

Independent hands-on reviews of opencodex were hard to find at the time of writing. The clearest public signal is adoption: the GitHub repository showed about 18,000 stars and 1,400 forks on November 2, 2026, five months after it was created, with 138 open issues. Aggregator GitTrend lists it among trending repositories. Several unrelated GitHub projects share the opencodex name, so make sure you install @bitkyc08/opencodex from the project linked on opencodex.me.

Pros

  • Free and MIT-licensed, with no account or hosted tier
  • Covers both Codex and Claude Code from one local port
  • 40+ built-in providers plus any OpenAI-compatible or local endpoint
  • Failover combos and account pools to stretch subscription quotas
  • Desktop app and web dashboard for people who prefer not to edit JSON

Cons

  • Desktop app is beta and the Windows installer is not code-signed yet
  • Dense, fast-changing docs; many options to configure
  • No independent reviews yet; adoption is the main public signal
  • Several unrelated projects share the opencodex name

Frequently asked questions

Is opencodex free?

Yes. opencodex is open source under the MIT License and costs nothing to install or run. You pay only for the models you route to, through your own API keys or provider subscriptions.

Can I use Claude Code with other models through opencodex?

Yes. opencodex serves the Anthropic Messages API on the same local port, and ocx claude launches Claude Code through it. Routed models from any configured provider then show up in Claude Code's /model picker.

Which providers does opencodex support?

The project lists 40+ built-in providers, including Anthropic, Google, xAI, Kimi, Ollama Cloud, Groq, OpenRouter, Azure, DeepSeek and GLM, plus any OpenAI-compatible endpoint and local providers.

How do I install opencodex?

Run npm install -g @bitkyc08/opencodex (Node 18 or later), then ocx init to add a provider and ocx start to run the proxy. Alternatively download the beta desktop app for macOS, Windows or Linux, or a standalone binary from GitHub releases.

Does opencodex work with the Codex desktop app?

Yes. opencodex writes a provider into your Codex config and its routed models appear in the Codex app model picker with reasoning effort levels, next to the native OpenAI models. ocx stop restores native Codex.

Can opencodex add costs I did not expect?

Only through providers you configure. The paid xAI image bridge is off by default, and combos only fall back to a pay-per-use target marked as metered when you explicitly allow it.

Compare Alternatives

See every opencodex alternative, compared

opencodex

Freemium · Web · API

Enjoying this? Get one like it in your inbox each morning.

one email a day · unsubscribe in two clicks · no third-party tracking