Grepture is an AI gateway that sits between your app and any LLM provider. However, you get to decide whether you add a proxy-hop and tinker with requests and responses in the hot path, or just track observability async in trace-mode.
Some of the highlights:
• Full traces of every request, including multi-step conversations
• LLM-as-a-judge evals running on live traffic and automatic dataset creation
• Automatic PII detection + redaction and rehydration before data hits the LLM
• Per-request cost and token analytics
• A zero-retention mode where prompt bodies never hit our DB (we're EU-hosted)
• CLI to pipe full dev sessions through the proxy
The proxy, CLI and SDK are open source already and the plan is to make the app available to self-host as well eventually. Right now we're focussed on making everything more useful for AI heavy products and optimize for developer experience.


