Your AI assistant is sharp in the moment and blank by morning. Neonmem gives it a memory that lives with your project — and the interesting part isn't just that it remembers, it's what it keeps and how it keeps it.

It remembers by kind, not as a transcript

Neonmem doesn't dump your chat history into a search box. It keeps your project as distinct kinds of memory — the way a good teammate carries a project in their head:

• Decisions — not just what you chose but why ("we went with Postgres because it scales under load"). The reasoning survives, not just the result.
• Dead-ends — the approaches that failed. Your agent stops re-suggesting the thing you already tried and threw out at 2am last Tuesday.
• Rules — your standing preferences and the project's invariants. "Always validate before saving." It just knows.
• Plans — where you're going, not only where you've been. The next steps and the goals stay in view.
• Questions, debates & observations — the loose threads and the context around them, so nothing important quietly evaporates.

Because each memory is a real, meaningful unit, your agent can reason over them — connect a decision to the dead-end that caused it, or a plan to the rule it has to respect — instead of just keyword-matching.

Memories live in zones — like a mind

Your memory isn't a flat list. It's organised into zones, the way thoughts settle in a brain:

• Reflex — the always-on core: the handful of things your agent should never forget, available instantly.
• Short-term — what you're working on right now, fresh and close at hand.
• Long-term — the settled history of the project, kept but out of the way until it's needed.

You can literally watch the zones glow and shift in a live 3D view as the memory grows — a brain filling in, session after session.

And then it dreams

This is the part people love. Like a mind, Neonmem sleeps and dreams. In a quiet consolidation pass it revisits everything it has gathered — strengthening what keeps proving useful, letting the noise fade, and finding connections between ideas it didn't notice in the moment. What started as scattered notes wakes up as understanding.

So the memory doesn't just grow — it matures. The longer you work together, the more it actually gets your project.

What that feels like, day to day

• You open a session and your agent already knows where you stopped — no re-explaining the project every morning.
• It catches "we tried that, it didn't work" before you lose the afternoon to it again.
• It works in your style, remembers your decisions, and keeps the plan in front of both of you.
• It's a colleague who was there yesterday — and last month.

All of it local, private, and yours: one file on your machine, no cloud, no third-party model reading your work.

source: https://neonmem.com/devlog/decisions-dead-ends-dreams

Try Neonmem · Windows & Linux · free for personal use