Hermes Atlas
Models, providers & proxies · works with Hermes Agent

codex-lb

Soju06/codex-lb

Load balancer and proxy that pools ChatGPT accounts behind OpenAI-compatible endpoints with a dashboard

In short

codex-lb is a load balancer and proxy that pools multiple ChatGPT accounts and exposes them through OpenAI-compatible endpoints, with usage tracking, per-key rate limits and a dashboard. Its client setup table lists Hermes Agent alongside Codex CLI, OpenCode and OpenClaw.

What codex-lb does

codex-lb load balances across several ChatGPT accounts and tracks per-account tokens, cost and 28-day trends in a web dashboard that supports password login with optional TOTP. API keys can carry rate limits by token, cost, window and model, and remote clients need a key created from the dashboard. Available models are synced automatically from upstream.

Clients connect to an OpenAI-compatible endpoint: http://127.0.0.1:2455/v1 for OpenCode, OpenClaw, Hermes Agent and the OpenAI Python SDK, and /backend-api/codex for Codex CLI and IDE extensions. The README recommends Docker, with uvx and Nix as alternatives, and stores data in SQLite by default, with PostgreSQL optional through CODEX_LB_DATABASE_URL. Per-client setup guides, including one for Hermes Agent, are on the documentation site.

Key features

  • Load balancing across multiple ChatGPT accounts
  • Per-account usage, cost and 28-day trend tracking
  • API keys with per-key limits by token, cost, window and model
  • Dashboard login with password and optional TOTP
  • OpenAI-compatible endpoints for Codex CLI, OpenCode, OpenClaw, Hermes Agent and the OpenAI SDK
  • SQLite by default, PostgreSQL optional

When to use it

  • Point Hermes Agent at the /v1 endpoint and share a pool of ChatGPT accounts
  • Watch per-account token usage and cost in one dashboard
  • Hand out limited API keys to remote clients

Who it is for: Developers who use several ChatGPT accounts with Codex-style tools and want pooling, quotas and usage tracking in one proxy.

How it fits with Hermes Agent

Lists Hermes Agent as a supported client with its own setup guide, using the http://127.0.0.1:2455/v1 endpoint.

How to install codex-lb

These commands are copied from the project's README. Check the repository for the latest steps before you run them.

uvx codex-lb
nix run github:Soju06/codex-lb

Requirements: ChatGPT accounts, added through the dashboard

FAQ

What is codex-lb?

codex-lb is a load balancer for ChatGPT accounts. It pools multiple accounts, tracks usage, manages API keys and shows everything in a dashboard.

Does codex-lb work with Hermes Agent?

Yes, the README lists Hermes Agent as a client using http://127.0.0.1:2455/v1 and links a Hermes Agent section in its client-setup guide. Remote clients need an API key created from the dashboard.

How do I install codex-lb?

The README recommends Docker and also shows uvx codex-lb and nix run github:Soju06/codex-lb. Then open localhost:2455 and add an account.

Similar models for Hermes Agent

All models
jundot oMLX

LLM inference server for Apple Silicon with continuous batching and SSD caching, run from the menu bar

★ 23kPythonModels ccusage ccusage

CLI that reports token usage and costs from local coding agent data, including Hermes Agent

★ 19kRustModels mnfst Manifest LLM Gateway

Open-source LLM gateway with one OpenAI-compatible endpoint, model routing, fallbacks and cost tracking

★ 7.6kTypeScriptModels lemony-ai cascadeflow

In-process runtime that routes agent model calls by cost, latency, quality and policy

★ 4.0kPythonModels Javis603 Token Monitor

Desktop widget that tracks token usage and costs across 43+ AI coding tools, including Hermes Agent

★ 2.7kJavaScriptModels xiufengsun TokenTracker

Local-first dashboard for AI token usage and cost across dozens of coding tools, with native apps

★ 2.0kJavaScriptModels

Related guides: How to install Hermes Agent