One OpenAI-compatible endpoint for all your coding agents. The routing policy lives in your account, not inside each tool. • Run DeepSeek 4.1 Flash (or other models) inside Codex, and keep your workflow • Give every API key its own daily budget, enforced per request • Explicit fallbacks: on a 5xx, 429 or timeout, try your next model • See remaining weekly quota and reset dates across Codex accounts Supports paid usage, or bring your own provider keys.
Hey Product Hunt 👋 I'm Idan, founder of Zerg.
ZergRouter started as an internal fix to a common problem. Once we were running more than one coding agent, every tool had its own provider setup, spend was scattered across keys, and when a Codex account hit its weekly limit mid-task, the only answer was "wait."
So we moved routing out of the tools and into one account:
• One OpenAI-compatible endpoint (zergrouter.com/v1), with a scoped key per tool
• Pick the model per client, with explicit fallbacks for when a provider has a bad minute
• A daily budget on each key, enforced per request instead of discovered on the invoice
• Codex can run open models like DeepSeek 4.1 Flash through the router, and you can see every Codex account's remaining quota and reset date in one place
For a limited time we are offering new accounts a 14-day DeepSeek trial; after that it's $5/mo plus usage, or bring your own provider keys.
This is the first of many upcoming launches, so would love your feedback. Have issues, this is our first launch - so please reach out and we'll be happy to support!
Dream