Edgee Codex Compressor V2
Product Hunt fmerian
Edgee says its gateway cuts Codex costs by 35.6% by compressing tokens first. It also reroutes and falls back automatically when providers slow down or hit limits.
Based on reporting by Product Hunt, fmerian — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Edgee is pitching a very specific kind of relief for coding agents: lower bills without forcing teams to change how they work. Its Codex Compressor V2 sits in front of Anthropic, OpenAI, and other LLM providers, compressing tokens before they go out, and it can also route requests to faster or cheaper models when asked. If a provider goes down, or a plan runs out, it falls back automatically. That’s the promise: the same Claude Code, the same Codex, just with less pain on the invoice.
The company says the new version brings a real drop in Codex spend. In a benchmark on the same repo, with the same model and the same workflow, Codex running through Edgee used 49.5% fewer input tokens. Cache hit rate moved from 76.1% to 85.4%. Total session cost fell by 35.6%.
That matters because token usage is the billable surface area of these tools, and coding agents can burn through context fast. Edgee’s pitch is not that it makes models smarter. It’s that it makes them cheaper to use by trimming what gets sent upstream and by squeezing more reuse out of cached context.
There’s also a team angle tucked into the product: attribution per repo and per PR. That’s the sort of feature that sounds boring until a company wants to know who used what, where, and why. Edgee is trying to sell infrastructure, not magic. And in a market full of grand claims, “same output, lower cost, no downtime” is a pretty practical line to hang a product on.
My take — AI-written commentary, not fact-checked reporting
This is the rare AI pitch that starts with a cost problem instead of another round of model worship. Good. Most teams do not need more theatrical intelligence; they need fewer surprise bills and fewer outages. The industry could use more products like this and fewer that sell you a bigger headache with a nicer demo.
Read more about this at: Product Hunt