andrsschultz cuts Mike's cord to Claude and Gemini

This fork rips out the hosted AI providers entirely and points Mike at a local model running on your own machine.

infrastructuresecurity

andrsschultz has stripped Claude and Gemini out of Mike and wired in a single local-AI backend instead - the kind you'd run on a workstation or an on-prem server with no internet round-trip. The bring-your-own-key screens, the model picker, the API-key prompts: all gone. There is one provider now, and it lives inside your network.

The plumbing underneath is careful work. The streaming code handles the chatty back-and-forth you need for tool use and for newer reasoning models that show their working. But the headline move is the air-gap: no client data leaves the building, and no per-seat AI bill to a US vendor. The trade-off is equally blunt - you are now responsible for the model, the hardware, and whatever quality gap exists between a local model and the frontier ones.

So what Worth a look for any firm or in-house team where data residency or confidentiality rules out hosted AI, and who has the infrastructure appetite to run their own.

View this fork on GitHub →

Spotted something wrong? Or know the PR text has fresher detail than the writeup above?

Commits in this thread

1 commit from andrsschultz/mike, oldest first. Source extracted verbatim from the harvested git log.

SHA Subject Author Date
5b79b283 Route LLM calls to local provider asschultz 2026-04-30 ↗ GitHub

Capture this thread into my fork

Download a single Markdown prompt that tells Claude how to port every commit above into your working tree — adapting paths and structure to match your repo. Run it via claude -p < capture-thread-82.md from inside the repo you want the changes in.

⬇ Download capture-thread-82.md