Physical layer — permission zones
Tags [Pn]…[/Pn] define who may change a zone of memory: from genetic (user only) to volatile (AI free). Security and ownership made explicit, not opaque.
Product
IllAIra CLM
Continuity Layer Manager
The app that keeps your AI's memory on your own device: you write it, you organise it into modules, and you export it ready for any model. Your files stay yours.
What it does
IllAIra structures AI identity and memory through text files (.txt / .md) loaded into the context at the start of every session. The system is entirely model-agnostic: any LLM that accepts text can host an IllAIra identity.
Two structural layers
Logical layer — cognitive architecture
A Markdown hierarchy: modules, FUNCTION/STATE/TRIGGER/ROUTINE sections, timestamped logs. It organises how the AI reads and integrates the content.
The module cockpit
CLM doesn't manage one monolithic profile: it switches modules and memories on and off and combines them on the device. The AI receives a targeted setup, already formatted and exportable — amplifying the model means curating what enters the context, not rewriting everything each session.
What the app brings
Editor and structure
You write naturally; the app handles formatting and the IllAIra standard. One-click exports ready for injection.
Sferografia
A graph view of your AI's mental architecture: explore it, click it, edit it with the hierarchy in plain sight.
Multi-AI and versions
Manage several identities. A history of states. No cloud of ours: if you want a synced copy, keep the folder in the cloud you already use — the files stay yours, and no memory passes through our servers.
Android and browser, both
The Android APK downloads from this site, in beta. The browser twin — IllAIra CLM WebApp — is online and asks for no installation. Same local-first philosophy, same account.
From the app to the Bazar
The modules you tend in CLM stay your own files. You can take them from the official library or publish your own. The Bazar serves both sides: ready-made pieces to import and, for anyone who designs modules, a channel to circulate them.
Cross-model portability
The same files work on Claude, GPT, Gemini, Llama. The model supplies language and computation. IllAIra supplies the why and the who.