Early release · open source
Bellamente
Long-term memory for AI agents that stays on your machine.
Bellamente gives agents durable, inspectable memory with semantic recall, versioned corrections, reversible forgetting, and a full audit trail.
Small software for real AI work.
We build tools for the parts of AI work that models alone do not solve: memory, handoffs, ownership, and evidence. The work is early, practical, and open enough to inspect.
Current projects
Open source · local-first
Agent Relay
A visible handoff layer for people and AI agents.
Agent Relay moves work between agents, tools, queues, and humans without losing the source, limits, status, or receipt.
Useful before impressive.
We start with a real job and ship the smallest thing that does it honestly.
Local and inspectable.
Your work should stay yours, and important system behavior should leave receipts.
Open where it matters.
Protocols, source, limitations, and release state should be visible enough to evaluate.
// learn from the work
Browse the guides →We publish the lessons too.
The free Starter Kit is our plain-English layer: practical guides for using AI, checking its work, and building something useful without swallowing the hype.