Top AI Product

As agents send longer and longer prompts, the token bill becomes the real cost of running AI in production. Edgee, an “agent gateway,” attacks that bill directly by compressing tokens before they ever reach the model provider.

## What Edgee does

Edgee sits between your app and the LLM as a transparent proxy. Its token compression runs at the edge on Rust-based infrastructure, shrinking prompt size by up to 50% while aiming to preserve semantic meaning and output quality. Because it works by swapping a single line — your base URL — there are no application code changes; you route requests through Edgee and they come out cheaper. It compresses both input and output tokens, which matters most for long-context apps, RAG-heavy workloads, and agentic loops where context windows fill up with redundant text.

## Turbo Models and fit

The latest piece, Edgee Turbo Models, targets the latency and setup pain of running open-weight models like GLM and Kimi K2.7, so teams can use cheaper open models without babysitting infrastructure. For anyone whose agent costs scale with token volume, a gateway that quietly halves prompt size is an infrastructure lever rather than a rewrite — which is the appeal of routing through it.