Trinity · In development · Web
NEXUS
The coordination hub for AI agents at work. Persistent project state, structured handoffs, and an audit trail on every recommendation. Built on the bet that the winning coordination layer is the most accountable one, not the smartest one.
What it is
Most teams running multiple AI agents lose context between them. Each session starts cold. Each agent forgets what the last one did. The coordination problem is not a model problem — it's an accountability problem.
NEXUS is the shared intelligence layer between every AI agent at the company. It structures the fragmented knowledge — Slack threads, email, docs, decisions, handoffs — into an executable project ontology that agents can run against. Every recommendation it makes carries a "why."
Customer agents stay with their providers — Claude, OpenAI, Cursor, whoever. NEXUS is the office those AI workers report to. Multi-vendor by design.
Who it's for
- Small businesses running real work across multiple AI agents and losing context between them.
- Operators who need a person in the loop and an audit trail behind every automated decision.
- Teams running parallel projects that share state but can't share a single agent without cross-contamination.
Status
Working prototype dogfooded on a real small business for about nine months. The MCP server is deployed and health-passing. Postgres backend with row-level security. Neural propagation layer plus a symbolic rule engine that attaches the audit trail to every recommendation.
Two reference spokes in development — EMAILY (email triage) and IAN (meeting intelligence) — to demonstrate the spoke pattern. NEXUS is the YC Summer 2026 application project for Trinity Studio.
Get on the waitlist
External-customer onboarding starts after the next batch of dogfood validation. Word-of-mouth waitlist open at the preview URL.