ConductLoop is a local-first "second brain" for people running multiple projects at once — a living memory graph plus AI agents that do work and write the result back into that memory. The hard constraint: everything private, running on the user's own computer, never on our servers. Here is how we shipped it.
The challenge
Most AI tools send your data to the cloud. ConductLoop's whole promise is the opposite: the memory graph, the local LLM and the agents all live on the buyer's machine. That meant packaging a full server into a desktop app and building a phone client that pairs to the user's own computer — no central backend.
What we built
- →Desktop app: a notarized Tauri build bundling a Python engine, memory graph and local LLM
- →iOS companion: a SwiftUI client that pairs to the user's own PC over the local network via QR
- →Licensing: a thin Cloudflare Worker + D1 license server so the paid app never exposes secrets
Result
A distributable, notarized desktop product with a paired mobile client — private by design, with zero data on our infrastructure. See it at conductloop.com.
Local-first is harder to build than a cloud SaaS, but for privacy-sensitive tools it is a genuine differentiator — and entirely achievable with the right architecture.
Work with us
Ready to build something?
Tell us about your project. We'll get back to you within 24 hours.
Start a project