Many sources. One living record.
AI memory that understands you.
Notes compact into living memory — then one clear answer.
Powered by a memory SLM · early access
Demo video coming soon
Notes → Compact → Reason → Answer — filmed walkthrough on the way.
Compare
The old way vs Archilas
Same problem. Two ways to handle memory.
The old way
Find bits of the past. Stuff them into the prompt. Hope it works.
- Search
Hunt for whatever seems relevant.
- Inject everything
Dump it all into the prompt.
- Token cost
That dump often burns a lot of tokens.
- No long-term reasoning
There’s no lasting picture to think with.
- No sense of time
What changed, what’s outdated, what still matters — unclear.
Search. Paste. Hope.
How Archilas is different
Keep a living record. Reason over it when you ask.
- Compact
Preferences, decisions, open loops — kept, not dumped.
- Reason
Thinks with that record when it supports the answer.
- Deliver
Clear answers into the tools you already use.
Compact. Reason. Deliver.
Built for Claude, ChatGPT, and Cursor. MCP coming soon — not live yet.
How it works
Notes. Compact. Reason. Answer.
Ship Friday?
Written as notes.
Why Archilas
Built around an SLM that reasons over compacted memory
In development / early access. Pre-launch — not live today.
Powered by a dedicated small model built specifically for memory — not a general chatbot model repurposed for recall.
Compact, purpose-built architecture — designed to reason over your history directly, not just search and paste text.
Works the same whether you're chatting or running an autonomous agent — one memory layer, every interface.
A small model built to reason over your compacted memory — so answers stay coherent without inventing bridges.
Surfaces
Built for the tools you already use.
Intended for Claude, ChatGPT, and Cursor. MCP support — coming soon. Not live integrations today.