🧠 how i work

Local AI, in the loop

Agents plan, patch and verify — intercepting failures before merge. Models run on my own silicon. Nothing leaves the building; nobody bills per token.

vs code claude code · cline agents plan · patch · verify intercept before merge llama.cpp local models · rtx 4090 · £0 api proxmox estate vms · lxc · gitops on-prem boundary

privacy by default

Prompts, code and logs stay inside the boundary. Compliance conversations get short.

agents with jobs

Not chat — agents that triage pipelines, write patches, and open PRs with receipts.

cost honesty

Inference on owned hardware. The meter reads zero; the trade-offs are documented.

See it live GitHub