Meet Lara
MEET LARA
What Lara Does
She Remembers. She Orchestrates. She Ships.
Lara holds the conversation. A team of specialist personas does the work behind her.
Remembers
Persistent memory built around your work — what you're shipping, who you're talking to, what you decided last week. She gets sharper the longer you use her, not duller.
Speaks
Real voice through the Brown Forces Voice engine — voice cloning, multi-language TTS, on-the-fly transcription. She talks back; she listens too.
Orchestrates
Hands off to specialist personas — a content reviewer, a research worker, future platform agents. Multi-agent collaboration, not one model trying to do everything.
Acts
Sixty-plus tools across vision, music, video, web search, email, schedules, content publishing, infrastructure health. She doesn't describe action — she takes it.
How It Works
Open the Playground. That's the App.
Lara runs on Kelex — a multi-tenant agent platform built on self-hosted brains and BFTS infrastructure.
Open her in the browser
No download, no install. The Kelex playground is the interface — chat, voice, file drop, live console for what she's doing in the background.
She learns your world
Memory writes when something matters and dedups when it doesn't. She reads from your projects, your schedule, your prior conversations — context every turn, not just within one session.
She delegates and ships
When work needs a specialist — a content reviewer, a research worker, a future platform agent — she routes it. Bounded-confidence flags surface only what she truly cannot decide alone.
Access
Currently in Private Alpha.
We're building Lara for daily use first; broader access opens once the core loop is steady.
Personal
- Full Kelex playground access
- Persistent memory + multi-persona routing
- All voice / vision / transcription / search tools
- Self-hosted brain (no third-party LLM)
- Bounded-confidence flag queue
Studio
- Your own Kelex tenant
- Custom personas for your team
- Your data isolated end-to-end
- Optional on-prem deployment
- Direct support from Brown Forces
Lara runs on self-hosted brains (Qwen3-Next-80B + GLM fallback) and the Brown Forces AI services (voice, vision, transcription, music, video). No cloud LLM lock-in.
For Businesses
Lara Shows Up in More Than One Product.
Lara is the intelligence inside Obsidian AI at Home — our residential AI appliance. Same persona, same voice, deployed on your own hardware behind your own walls.
See Obsidian AI at Home→She's also the voice of the K-2 Math Screener, reading questions aloud to kindergarteners in their native language.
Learn More→Want to Try Her?
Alpha access is invite-only while we tune the loop. Tell us what you'd use her for.