## Summary
sure-scale/mike
A privacy-minded take on Mike that can route model calls through a gateway which pseudonymizes prompts before they reach the LLM providers.
This is sure-scale's fork of Mike, and so far it tracks upstream closely - the interesting part is a direction rather than a shipped product. The work here, proposed by @nicksolarsoul, is an optional routing layer that sends Mike's model calls through a third-party gateway called Hey Jude, which pseudonymizes prompts before they reach the downstream providers (Claude, Gemini, OpenAI).
The design keeps it conservative: the behavior lives behind a single enable flag and is off by default, so anyone running this fork gets ordinary direct-to-provider calls unless they deliberately turn the gateway on. The stated motivation is privacy - stripping identifying details out of prompts before they leave your environment.
If you're a legal-tech reader weighing whether client data should ever hit a model provider in the clear, this is the fork exploring an answer. It's early, and it reads as a proposal more than a finished feature, but the intent is clear enough to be worth a click through to GitHub.
What's in it
- Pseudonymizing gateway option Model calls can be routed through Hey Jude, a gateway that scrubs identifying details out of prompts before they reach the LLM providers.
- Off by default The whole behavior sits behind one enable flag and stays off unless you opt in, so existing direct-to-provider calls keep working unchanged.
- Covers the major providers The routing is aimed at Mike's Claude, Gemini, and OpenAI calls rather than a single vendor.
- Privacy-first framing The reason to reach for it is keeping sensitive prompt content out of the clear on its way to a model - a natural fit for client-confidential work.
Direction
securityredactionintegration
Activity
## Summary
📝 nicksolarsoul floats a privacy gateway for Mike's AI calls, then pulls it 0 commits — securityinfrastructure draft
A proposal to scrub client names and organizations out of prompts before they ever reach the AI providers - opened, and closed again within a minute.
📝 nicksolarsoul pitches a privacy gateway that scrubs names before they hit the AI 0 commits — securityintegration draft
An opt-in routing layer would strip identifying details out of prompts before Mike sends them to Anthropic, Google, or OpenAI.
Threads of work (detailed view)
nicksolarsoul pitches a privacy gateway that scrubs names before they hit the AI
An opt-in routing layer would strip identifying details out of prompts before Mike sends them to Anthropic, Google, or OpenAI.
nicksolarsoul floats a privacy gateway for Mike's AI calls, then pulls it
A proposal to scrub client names and organizations out of prompts before they ever reach the AI providers - opened, and closed again within a minute.
Pull requests (detailed view)
⛔ Closed without merge (2)
by
nicksolarsoul · opened 2mo ago · closed 21d ago by
nicksolarsoul · opened 2mo ago · closed 2mo ago