A model answers. An engine orchestrates, remembers, and stays yours.
The Conductor plans your task and hands each part to the model best at it — Opus for the hard reasoning, Gemini Flash for the fast work — running independent steps in parallel across git worktrees and cross-reviewing the code between providers.
Learn more →Memory, a learned machine profile, your recorded skills, and a graph of your codebase — fed into every session so agents start knowing your repo instead of re-exploring it. Recall stays lean: only the files that actually match, and only the relevant slice of each, so context never balloons your token bill.
Learn more →Every session executes locally and your keys never leave — the cloud only ever sees counts and ciphertext. Linux-first down to real desktop control, with Windows and macOS builds rolling off public CI.
Learn more →One engine. Every surface you work from.
Everything below ships today — the same local engine and the same persistent brain, whether you drive it from a window, a terminal, or a service.
Connect the AI
you already pay for.
No new subscription. Point Lectern at Claude Code or Antigravity — it routes each task to the model that's best at it, and your memory and skills follow across both.
Built in public, licensed to stay open.
Lectern is Apache-2.0 — a local-first tool you can audit, fork, and help build. Stars fuel the roadmap; issues steer it.
Built for Linux, first.
A real native package — not a Mac port. v0.8.0 · Jul 11, 2026. Bring your own backend; keep your memory forever.
Start the engine.
Free to download. Bring your own backend. Keep your memory forever.