We built our own infrastructure. In the open.
When an engagement needs memory, observability, or an agent runtime, we reach for tools we wrote and maintain, not a vendor you will be renting in three years. These ship today, and they are the clearest proof of what we can build for you.
Agent memory · MCP-native
Liveextremis
A memory platform for AI agents with episodic, semantic, procedural, and identity layers, exposed over MCP.
This is the memory layer behind our organizational-memory builds.
LLM observability · on PyPI
Livepeekr
LLM observability and tracing. Auto-instruments OpenAI, Anthropic, and Bedrock with a couple of lines.
This is the tracing layer behind our LLM-observability and incident builds.
AI assistant · research
ResearchFriday
A personal AI assistant research project: BDI reasoning and a persistent agent loop.
Where we work out the agent patterns that later harden into client builds.