Universal provider proxy for OpenAI Codex, Claude Code, Claude Desktop & Grok Build
Two commands, and every one of them runs any LLM you point it at.
bun install
bun run build:gui
bun run src/cli/index.ts start
English · 한국어 · 简体中文 · Русский · 日本語 · 📖 Full documentation →
CodexCommander is a lightweight local proxy that translates Codex's Responses API into whatever your provider speaks — streaming, tool calls, reasoning tokens, images, in both directions. Use Claude, Gemini, Grok, GLM, DeepSeek, Kimi, Qwen, Ollama, or any other LLM with Codex, Claude Code, Claude Desktop, and Grok Build. It can also manage a ChatGPT account pool for Codex auth: add accounts, refresh their quotas in the dashboard, and let new sessions auto-route to the lowest-usage healthy account while existing threads stay pinned to the account that started them.
bun install
bun run build:gui
bun run src/cli/index.ts start # or use `service` instead of `start`The registry package is not currently published. These instructions run the checked-out source
with a locally installed Bun. In this checkout, any documented ccx <args>
command can be run as bun run src/cli/index.ts <args>.
Open http://localhost:10100 and configure everything in the web dashboard — add providers
(40+ built-ins, or any OpenAI-compatible endpoint), pick models, manage accounts. ccx gui
re-opens the dashboard at any time.
It can also manage a ChatGPT account pool for Codex auth. Add multiple ChatGPT / Codex accounts,
refresh their 5h / weekly / 30d quota in the dashboard. Under quota routing, new sessions can use
the lowest-usage healthy account; round-robin and fill-first use their own policies. Existing Codex
threads normally retain affinity to the account that started them, so long SSH, tmux, or
mobile-connected sessions do not jump accounts mid-conversation — but quota re-evaluation, failover,
account exclusion, affinity expiry, or 401/403 and 429 recovery can rebind them. Give the accounts a
selection order when one of them — usually your Codex Desktop login — should only be reached for
once the others are drained.
The native companion shows live agent activity and provider quota windows, then opens the existing dashboard for OAuth, API keys, account management, and logs. It is a UI layer over the running proxy: it does not use Keychain, create a second provider-login system, or store provider credentials in the app.
No packaged macOS app is currently published. Build the companion from this checkout and keep that
one development app at dist/macos/CodexCommander.app instead of copying it to Application Support:
bun install
bun run test:macos
bun run build:macos
open dist/macos/CodexCommander.appEvery built app launches only the Bun runtime and server resources embedded in its own
Contents/Resources/runtime; it never executes checkout src/ or an ambient ccx. Rebuild the app
to pick up source changes. If startup fails, the menu app stays open so its diagnostics and Start
control remain available. Quit closes only the companion UI. Stop and Restart are
separate, confirmation-gated proxy actions.
On its first launch from dist/macos or Applications, the app enables Launch at Login so the
menu icon returns after sign-in. The startup row exposes the actual mode: Desktop launches the
menu app, Headless leaves only an installed background service at login, and Off starts
neither automatically. Rebuilt source apps refresh their login registration in place; they are
never copied into Application Support. Full
setup, Gatekeeper, release packaging, and troubleshooting details are in the
macOS menu bar guide.
ccx opencode starts OpenCode with a transient, in-memory provider.codexcommander block and never
rewrites OpenCode config files. For plain OpenCode or the Desktop app, use the dashboard's
Integrations page to apply the opt-in, reversible connection that owns only
provider.codexcommander; see the OpenCode guide.
bun run src/cli/index.ts start # or use `service`
bun run src/cli/index.ts init # interactive setup: writes ~/.codexcommander/config.json and wires Codexccx init never starts the proxy; start it first (or after — either order works, but headless
commands like ccx provider add and ccx combo set talk to the live proxy and exit nonzero
when it is unreachable). ccx status / ccx doctor / ccx health report the running state.
| OS | Status | Service manager |
|---|---|---|
| macOS (arm64 / x64) | Fully supported | launchd |
| Linux (x64 / arm64) | Fully supported | systemd (user unit) |
| Windows (x64) | Fully supported | Task Scheduler (hidden) / opt-in native service (--native, WinSW) |
Running from source requires Bun. Windows runs natively without WSL. See the installation docs.
- Use any LLM with Codex, Claude Code, Claude Desktop, and Grok Build — 40+ providers out of the box, each keeping its own native UI.
- Pool ChatGPT accounts safely — thread affinity, quota-aware auto-switching, cooldown and fail-closed auth handling.
- Combos — one virtual model id with failover or weighted round-robin across providers. See the combo guide.
- Sub-agents on any model — feature routed models in Codex's sub-agent picker, with v1/v2 surface control and fallback chains. See the sub-agent guide.
- Log in once, skip the API key — OAuth for xAI, Anthropic, and Kimi; or forward
codex login, paste a key, or use${ENV_VAR}references. - Web search & vision sidecars — non-OpenAI models get real web search and image understanding through a sidecar over your ChatGPT login.
- See what's happening — the dashboard shows providers, OAuth status, model selection, and a live request log with cache token counts.
- Native macOS glance view — see active parent/subagent rows and every available session, weekly, monthly, or credit quota window without duplicating provider setup outside the dashboard.
- Clean exit, zero residue —
ccx stoprestores Codex to its original configuration.
Target any configured provider and model with the provider/model syntax:
codex -m "anthropic/claude-opus-5" "Explain this stack trace"
codex -m "google/gemini-3-pro" "Write unit tests for auth.ts"
codex -m "ollama/llama3" "Refactor this function"Omit the provider/ prefix to use the default provider or auto-match by model name pattern.
Provider model ids containing / are exposed with inner slashes aliased to -; the raw
full-slash form keeps working too. Details: model routing docs.
OpenAI (ChatGPT login or API key), Anthropic, Google Gemini, xAI, Kimi, Azure OpenAI, Ollama
(local + Cloud), Cursor (experimental), and every OpenAI-compatible endpoint — plus DeepSeek,
Groq, OpenRouter, Together, Fireworks, Cerebras, Mistral, Hugging Face, NVIDIA NIM, MiniMax,
Qwen Cloud, SiliconFlow, and more. Full list: ccx init or the
provider docs.
ccx init # interactive setup (writes config, wires Codex, offers the shim)
ccx start [--port 10100] # start the proxy in the foreground
ccx stop # stop + restore native Codex
ccx service [install|start|stop|status|uninstall|remove] # background service
ccx codex-shim install # start the proxy on demand whenever `codex` launches
ccx health [--json] # check immediate proxy liveness
ccx ready [--json] [--wait [--timeout <seconds>]] # check post-sync readiness
ccx status # is the proxy running?
ccx gui # open the web dashboard
ccx provider <...> # manage providers (list/add/edit/test/remove)
ccx account <...> # manage ChatGPT accounts & API-key pools
ccx combo <...> # manage failover / round-robin combos
ccx v2 <...> # multi-agent v1/v2 surface controlsUnpinned starts may pick another free port if the preferred one is busy; an explicit --port
never hops. Full reference: CLI docs.
GET /healthz reports immediate proxy liveness. The unauthenticated GET /readyz endpoint reports
post-sync readiness with the sanitized JSON identity {service, version, uptime, pid, port, status}.
It returns 200 when status is ready; pending and terminal failed return 503 with
Retry-After: 1.
ccx ready [--json] [--wait [--timeout <seconds>]] performs one probe by default. --wait polls
for up to 45 seconds by default, but exits immediately when it observes terminal failed;
--timeout <seconds> sets a 1–300 second limit, requires --wait, and accepts only positive integers. CLI --json output is
{ready, status, pid, port}, where status is ready, pending, failed, or unreachable.
| Exit | Result |
|---|---|
0 |
Ready |
1 |
Not ready: pending, failed, timeout, or unreachable |
64 |
Invalid arguments |
Use the service (ccx service) for an always-on proxy that restarts on crash. Use the
shim (ccx codex-shim install) for lightweight, on-demand startup without a background
daemon. Remove them with ccx service uninstall / ccx codex-shim uninstall.
ccx uninstall # stop, remove service/shim, restore native Codex, clean up stateBy default CodexCommander binds to 127.0.0.1 and needs no extra authentication. Binding beyond
loopback ("hostname": "0.0.0.0") requires a bearer token — the proxy refuses to start
without CODEXCOMMANDER_API_AUTH_TOKEN, and every client request must carry it as
x-codexcommander-api-key. Details: configuration reference.
The docs — install, providers, routing, combos, sub-agents, sidecars, integrations, and the
CLI/config/management-API references — live in docs-site/. They build locally;
publishing automation is not included in this repository.
Maintainer source-of-truth notes live under structure/, contributor setup in
CONTRIBUTING.md, and security reporting in SECURITY.md.
Report undisclosed vulnerabilities privately through
GitHub private vulnerability reporting,
not a public issue.
Source development requires the bun CLI on your PATH.
cd /path/to/CodexCommander
bun install
bun run typecheck
bun run testSee Contributing.
CodexCommander is an independent, community-maintained project and is not affiliated with or endorsed by OpenAI, Anthropic, or any other provider.
Some providers — notably Anthropic (Claude) — may suspend or restrict accounts that route API traffic through third-party proxies. Use at your own risk (UAYOR). Before connecting a provider, review its Terms of Service to confirm that proxy-based access is permitted. The CodexCommander maintainers are not responsible for any account actions taken by upstream providers.
MIT
The MIT license permits forks, modification, distribution, and sale subject to its notice condition. It does not grant third-party product names, logos, or trademark rights. A rebrander or commercial redistributor must audit names, logos, and trademarks; preserve the MIT and applicable third-party notices; avoid claims that it pools or resells third-party credentials or subscriptions; and use its own bundle identifier, signing and notarization, distribution metadata, and package metadata. This is a release-operational checklist, not legal advice.