opencodex

Claude Desktop, running any model

Opus answers, then hands the task to a GPT-5.6 Sol subagent.

Grok Build, running any model

Sol drives the session and calls a Kimi K3 subagent.

English · Français · 한국어 · 简体中文 · 繁體中文 · Русский · 日本語 · Türkçe · 📖 Full documentation →

opencodex 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.

Quick start

Personal install (CLI)

npm install -g @bitkyc08/opencodex   # Node 18+; the Bun runtime is bundled automatically
ocx start                         # proxy + dashboard on localhost:10100

Use ocx service to run it in the background.

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. ocx gui re-opens the dashboard at any time.

Desktop app (beta)

The desktop app is the same proxy and dashboard in a native window, with a tray and bundled ocx. It attaches to a proxy that is already running, or starts its bundled one, and the dashboard stays on the proxy port (http://localhost:10100 unless you configured another). Pick the file for your platform from the latest release:

Every file has a .sha256 next to it on the release page. On macOS 14+ the app also ships a WidgetKit extension that shows proxy status, today's usage and provider quotas; the snapshot model it renders lives in app/ (MenuBarCore). To build the app yourself, run bun install && bun run build:gui at the repository root, then in desktop/ run bun install && bun run prepare-sidecar && bun run prepare-widget && bun run build:local on macOS, or bun install && bun run prepare-sidecar && bun run build:local on Windows and Linux (the widget step needs macOS). The Desktop App guide and the macOS Menu Bar App guide cover first launch, and AGENTS_INSTALL.md lists everything written to disk.

ChatGPT account pool

opencodex 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.

Sponsors

Sponsors keep opencodex maintained across every upstream protocol change. Interested? See SPONSORS.md.

Thanks to OrcaRouter for sponsoring this project! OrcaRouter is one OpenAI-compatible AI gateway for production AI: adaptive routing that grades every prompt and sends it to the model that clears your bar, automatic failover, routing rules as code, zero-markup provider pricing with prompt caching, and guardrails, an agent firewall, and request logs on every call across 200+ models. Pick OrcaRouter in the Add provider picker or run ocx provider add orcarouter; orcarouter/auto is the adaptive router.

Thanks to TokenLab for sponsoring this project! TokenLab gives coding agents one API key for leading models, supporting OpenAI Responses and Chat Completions, Anthropic Messages, and Gemini's native API formats, with streaming and tool calling. It also provides an MCP server and agent Skills for easy integration. Choose your delivery mode and pay as you go. Pick TokenLab in the Add provider picker or run ocx provider add tokenlab.

TokenLab 为编程智能体提供统一的多模型 API，一枚 API Key 即可接入主流模型，支持 OpenAI Responses、Chat Completions、Anthropic Messages 和 Gemini 原生 API 格式，以及流式输出和工具调用。同时提供 MCP 服务器和 Agent Skills，方便接入现有工作流；交付模式可选，按量付费。

Docker Compose

The repository ships a digest-pinned, non-root Compose build. The build generates and verifies the canonical compatibility manifest from the selected Git snapshot. A local clone needs Git and Docker Compose; a remote Git context needs Docker Compose. Neither path needs host Bun or a preparation step. Initialize the data-plane token once through stdin and start the hub:

git clone https://github.com/lidge-jun/opencodex.git
cd opencodex
docker compose build
openssl rand -hex 32 | docker compose run --rm -T hub bun run docker/bootstrap-token.ts
docker compose up -d
curl --fail --silent http://127.0.0.1:10100/healthz
curl --fail --silent http://127.0.0.1:10100/readyz

The default host binding is 127.0.0.1:10100. Remote exposure requires explicit OPENCODEX_BIND_ADDRESS= docker compose up -d; 0.0.0.0 opts into all host interfaces. Restrict access with a firewall and an authenticated TLS/tailnet frontend. The generated JSON stays untracked. The build context admits only .git/index and .git/HEAD — the inventory git ls-files reads, about 1 MB rather than the full object store — and they are visible only to the build-only manifest stage through a read-only mount, so no COPY includes .git. An existing host-generated manifest is still accepted only after validation; otherwise the build generates one itself. The build rejects stale manifests, missing or mismatched files, extra source files, and symlinks. It checks every recorded SHA-256 against the build context and copied runtime files, including package.json, bun.lock, and the specifically included scripts/model-metadata.source.json.

A remote Git context needs BuildKit to retain Git metadata. This Compose build fragment selects the remote snapshot and passes the required built-in argument:

services:
  hub:
    pull_policy: build
    build:
      context: https://github.com/lidge-jun/opencodex.git#main
      dockerfile: Dockerfile
      target: runtime
      args:
        BUILDKIT_CONTEXT_KEEP_GIT_DIR: "1"

The token and mutable state stay in the ocx-state named volume; no credential is placed in the image, Compose file, environment, or shell arguments. See the Remote Hub deployment guide for provider setup, authenticated acceptance checks, remote management, and rollback.

Install from source (latest dev)

macOS / Linux:

curl -fsSL https://bun.sh/install | bash
git clone -b dev https://github.com/lidge-jun/opencodex.git
cd opencodex && ~/.bun/bin/bun install
~/.bun/bin/bun run build:gui
~/.bun/bin/bun run src/cli/index.ts start

Windows (PowerShell):

irm bun.sh/install.ps1 | iex
git clone -b dev https://github.com/lidge-jun/opencodex.git
cd opencodex; bun install
bun run build:gui
bun run src/cli/index.ts start

Source install runs the latest dev branch. Memory ownership patches, runtime GC improvements, and unreleased fixes are available here before they reach the npm package.

For agents

npm install -g @bitkyc08/opencodex
ocx start     # or `ocx service`
ocx init      # interactive setup: writes ~/.opencodex/config.json and wires Codex

ocx init never starts the proxy; start it first (or after — either order works, but headless commands like ocx provider add and ocx combo set talk to the live proxy and exit nonzero when it is unreachable). ocx status / ocx doctor / ocx health report the running state.

Agents installing or running opencodex: read AGENTS_INSTALL.md. An interactive ocx start may ask once whether to star this repository — that is the user's decision, never an agent's. The CLI suppresses the prompt for agent-driven runs and the API refuses them with 403 agent_consent_required.

Supported platforms

The CLI install requires Node 18+; the desktop app needs neither Node nor Bun. The Bun runtime is bundled on npm install — no separate Bun install needed, no WSL needed on Windows. If npm blocked the bundled runtime's install scripts, see the installation docs.

Highlights

• 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 — thread affinity, quota-aware auto-switching, cooldown and fail-closed auth handling.

  Provider-policy note: Account pooling is for routing and operational resilience only; it does not guarantee protection from provider rate limits, enforcement, suspension, or other account actions. OpenCodex does not endorse using additional accounts to circumvent provider limits or sharing account credentials between people. You are responsible for complying with each provider's current terms. See the Codex Auth account-pool guidance and OpenAI's current Terms of Use.

• 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.

• Clean exit, zero residue — ocx stop restores Codex to its original configuration.

• Bounded memory ownership — every long-lived cache, ring buffer, and protocol-translation store has a finite cap, byte budget, or active reconciliation. No unbounded Map or Set survives a config reload.

Memory ownership details

OpenCodex tracks process-retained state in the categories below. Each has a documented bound:

• 14 retained stores (request log, debug rings, image cache, model cache, vision descriptions, cursor blobs, responses continuation, etc.) are byte-accounted and evicted by the app-owned memory budget (default 256 MiB), except the native control replay store, which is pinned and never evicted.
• 4 observed buffers (translator accumulators, image/OAuth/Grok tails) are monitored for in-flight byte pressure without eviction.
• 31 state-store registrations handle expiry sweeps (60 s interval) and config-generation reconciliation so stale provider/account keys are removed.
• Path and fingerprint memos (workspace metadata, hardened identities, installation salts, mode-hint capabilities) use insertion-order LRU caps (8–128 entries).
• Model-cache generation tombstones are deleted after reconciliation; a global generation increment prevents stale in-flight discoveries from repopulating removed providers.
• Lab event-id deduplication runs under a ledger lock from disk, with no process-level RAM index.

Run GET /api/system/memory (with the admin token) to inspect live retained bytes, eviction counters, and watchdog samples.

Model routing

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.

JEV Auto routing (optional)

TypeSafe JEV can choose the first model and reasoning effort for an opt-in Combo while the normal model picker and every direct route stay unchanged. Add the credential with ocx login jev, from Providers → TypeSafe JEV → Add API key, or through TYPESAFE_API_KEY/JEV_API_KEY. Then open Models → Combos → Create JEV Auto, choose the allowed target models, and check the exact efforts JEV may select for each target. Leaving a target's effort setting untouched allows all efforts that model currently advertises.

JEV is consulted only for jev-auto and only once per logical model call. Missing credentials, network failures, or invalid decisions fail open to the first currently eligible target; caller cancellation still cancels the request. Automated tests use a mocked TypeSafe endpoint and do not validate a live JEV account.

A JEV Combo can instead ask a self-hosted decision model, such as Ollama's keyless tev1: add a jev-decision provider whose baseUrl is the full /v1/systemone endpoint and set the Combo's decisionProvider to it. TypeSafe credentials are never sent there. Details: System One-compatible server.

Providers & adapters

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, OpenGateway, Together, Fireworks, Cerebras, Mistral, Hugging Face, NVIDIA NIM, MiniMax, Qwen Cloud, Qoder Global and CN (official PAT + CLI), SiliconFlow, and more. Full list: ocx init or the provider docs.

CLI

ocx init                       # interactive setup (writes config, wires Codex, offers the shim)
ocx start [--port 10100] [--socks5 [host:port] | --socks5-off]  # SOCKS5 defaults to socks5://127.0.0.1:10808
ocx stop                       # stop + restore native Codex
ocx service [install|repair|restart|start|stop|status|uninstall|remove]  # background service
ocx codex-shim install         # start the proxy on demand whenever `codex` launches
ocx health [--json]            # check immediate proxy liveness
ocx ready [--json] [--wait [--timeout <seconds>]]  # check post-sync readiness
ocx status                     # is the proxy running?
ocx gui                        # open the web dashboard
ocx provider <...>             # manage providers (list/add/edit/test/remove)
ocx account <...>              # manage ChatGPT accounts & API-key pools
ocx combo <...>                # manage failover / round-robin combos
ocx v2 <...>                   # multi-agent v1/v2 surface controls
ocx update [--tag preview]     # update opencodex

A start whose preferred port is busy stops and names the holder instead of moving to another port, so it can never leave a second proxy running beside the first. Free the port, or name a different one with --port. Full reference: CLI docs.

Health and readiness

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.

ocx ready [--json] [--wait [--timeout ]] performs one probe by default. --wait polls for up to 45 seconds by default, but exits immediately when it observes terminal failed; --timeout 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.

An older proxy without /readyz fails closed as unreachable with exit 1, while ocx health remains compatible.

Autostart: service vs shim

Use the service (ocx service) for an always-on proxy that restarts on crash. Use the shim (ocx codex-shim install) for lightweight, on-demand startup without a background daemon. Remove them with ocx service uninstall / ocx codex-shim uninstall.

Uninstall

ocx uninstall                  # stop, remove service/shim, restore native Codex, clean up state
npm uninstall -g @bitkyc08/opencodex

Remote access

By default opencodex 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 OPENCODEX_API_AUTH_TOKEN, and every client request must carry it as x-opencodex-api-key. Details: configuration reference.

Documentation

The public docs — install, providers, routing, combos, sub-agents, sidecars, integrations, and the CLI/config/management-API references — are built from docs-site/ and published to opencodex.me.

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. That form is the only technical channel — there is no security email. Follow-ups stay in the private report itself; a public issue may carry coordination only, never vulnerability details. Acknowledging a report is not the same as triaging it, and no first-response target is promised.

Development

Source development requires the bun CLI on your PATH. The published npm package bundles its own Bun runtime for installed ocx commands; package scripts may also resolve Bun through that bundled dependency.

git clone https://github.com/lidge-jun/opencodex.git
cd opencodex
bun install
bun run typecheck
bun run test

See Contributing.

Contributor work that landed through a maintainer carry or reimplementation, where the commit does not name its original author, is recorded in CREDITS.md.

Code signing policy

Windows binaries are moving to Authenticode signing (#7207). The Windows Code Signing guide covers the rollout, how to verify a signature, and the full policy.

Free code signing provided by SignPath.io, certificate by SignPath Foundation.

• Committers and reviewers: lidge-jun
• Approvers: lidge-jun
• Privacy: This program will not transfer any information to other networked systems unless specifically requested by the user or the person installing or operating it.

Disclaimer

opencodex 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 opencodex maintainers are not responsible for any account actions taken by upstream providers.

License

MIT

内容由编辑整理，部分信息来自 GitHub 与项目官网，版权归原作者所有。发现错误？报告问题