A continuous multimodal sensory-state layer for persistent AI agents. Environment in, agent-readable perception out — and music is where we prove it in public.
Persistent AI agents are blind and deaf between prompts. They reason in text, but the world they operate in is continuous — sound, ambience, state, rhythm. Today's agents get a snapshot when a user happens to type something. There is no general layer that turns a continuous environment into a compact, persistent, agent-readable sensory state.
Perception work in the industry is mostly single-shot: transcribe this, classify that. What's missing is ongoing perceptual state — compressed, persistent, queryable, and able to trigger cognition only when something actually changes.
Auris is a sensing pipeline that runs continuously at the edge and maintains a living perceptual state an agent can read, remember, and act on:
The agent doesn't poll the world. The world accumulates meaning in the background, and the agent wakes up when the state says something happened. That's the difference between a chatbot and something that is present.
Compression with memory, not transcription without context. Auris doesn't produce a stream of labels; it maintains an evolving state object with history, so "this changed" is a first-class signal. It's built by the same team that built BCC (patent-pending AI memory infrastructure) — persistent state is our home turf. And it's designed to plug into agent ecosystems (MCP-connected sensory systems) rather than sit inside one vendor's walled garden.
Your music isn't just sound. It's a map of how you feel — one of the most honest signals of inner emotional state there is. When you're grieving, you don't search "sad songs" — you gravitate toward them. Auris listens to what you listen to and maps each song across eight dimensions of human feeling, building a memory of your sonic identity over time: the patterns, the peaks, the moments when a song hit you so hard it changed your afternoon.
Every song gets mapped across these axes. Together they form a unique emotional fingerprint — no two songs feel exactly the same, and no two people hear the same song identically.
Paste a YouTube link, or let Auris detect what's playing. It works with any audio source — no platform lock-in.
Each track is analyzed across all eight dimensions and placed into your personal affect space. You get a color orb showing the emotional signature — warm, cool, intense, gentle — at a glance.
Every song becomes a point in your timeline. Over weeks and months, patterns emerge: the songs that pulled you through hard days, the albums that defined a season, the peaks where music became transcendent.
Full-text search across your music memory. Find "that warm song from last Tuesday" or "anything with high tension and low warmth." Your memory, searchable by emotion.
Streaming platforms reduce your emotional life to a "liked songs" playlist and an engagement algorithm. Auris doesn't judge what you listen to and doesn't sell recommendations. It helps you see what your music already knows about you.
For people: anyone who feels music deeply and wants to understand why. Musicians mapping creative influences. Therapists exploring emotional states through listening patterns. Writers building playlists for characters.
For builders and partners: teams working on persistent agents, robotics, ambient computing, games, or adaptive interfaces who need a sensory-state layer that already runs. Auris is licensable, pilotable, and weird in the ways that matter.
Auris is included with every Reflexion subscription. Commercial licensing and pilots are scoped per engagement.