Hey Product Hunt,

Codex Pooler is a self-hosted gateway for expert teams that need to share and operate Codex account capacity as infrastructure.

Instead of binding every agent, CLI, or internal tool to one account, you add Codex accounts to Pools, issue stable Pool API keys, and let the gateway route work by model support, limits, health, policy, and session continuity.

What pooling gives you:

- one stable key representing many Codex accounts

- central capacity changes without reconfiguring every client

- fewer blocked workflows when one account is exhausted or unhealthy

- long-running sessions and reconnects staying on the right upstream

- clearer visibility into which account handled a request, and why

Works with multiple harnesses:

- Codex CLI

- OpenCode

- editor agents

- internal automations

- custom agent workflows

Operator layer:

- alerts and incidents

- stats, request history, routing views, and audits

- Prometheus metrics and a Grafana dashboards

- advanced tuning for limits, diagnostics, circuits, ingress, model metadata, and timeouts

Deployment:

- self-hosted

- Docker Compose for trying it

- Helm for Kubernetes with separate app, worker, scheduler, and migration roles

The goal is simple: pool Codex capacity, keep sessions reliable, and give teams enough control to run Codex-powered work seriously.

Docs: https://docs.codex-pooler.com

Code: https://github.com/icoretech/codex-pooler

I'd love feedback from teams already juggling multiple Codex accounts. Where would pooling help first, and what controls would you need before trusting it?