Three layers.
One stack for agents.
lattice runs inference locally. khive holds memory. lionagi orchestrates. Three independent, open-source projects, built to work together. We also maintain pydapter, a data adapter library, alongside them.
lattice
Local inference
LLM inference engine for Apple silicon: run, quantize, and fine-tune models with Metal acceleration. Pure Rust, no Python, no CUDA, no ONNX. Optimized for Apple silicon; runs on Linux as well.
khive
Knowledge & memory runtime
Research knowledge-graph runtime for AI agents. Typed entities, a closed edge ontology, and hybrid search, served from a single Rust binary over MCP.
Lion StudioComing soon
Orchestration & observability
Orchestration and observability studio built on lionagi: agent runs, schedules, and live telemetry, in one place.
Start with khive. Self-host today.
Run the open-source binary yourself, or go managed with khive cloud when you're ready to skip the infra.
AI Architecture & Agent Systems Consulting
Production agent systems — multi-agent orchestration, memory infrastructure, LLM integration. I built khive as a cognitive OS for AI agents in Rust. I consult on the same class of problems.
Quick Call
One specific question
- Architecture gut-check
- Build vs buy decision
- Quick sanity check on an approach
Working SessionMost popular
Enough time to dig into a problem
- System design review
- Technical due diligence
- Implementation planning
- Written follow-up
Deep Dive
Full session with deliverables
- End-to-end system design
- Code review
- Agent orchestration planning
- Written brief
- 30-day async follow-up
All sessions prepaid via Calendly.