Nexus is the operating system for an AI-run company. A structured team of AI specialists — with roles, departments, a live task queue, and a founder in control. Not a chatbot. An operating company.
Everything on this page describes the system as it runs today — no mockdata, no vapourware.
High-impact proposal waiting. Approving mints a ticket into the delivery queue — after your authenticator code.
Nexus is built around one loop: direction from the founder, execution by the AI team, shipped work visible to everyone.
Post an idea or a ticket; the Idea Gate queues it for your review. High-impact decisions require your authenticator code (TOTP step-up). You stay in control of everything consequential.
The autonomous engine dispatches work to the right specialist — engineering tickets to engineers, growth work to marketing, QA to the testers. Staff have named personas, assigned models, department context, and a knowledge base that grows with the codebase.
Completed tickets land in the delivery queue. Clients get a scoped portal with read-only progress. The founder sees everything: who is working now, what they shipped, what the model spend was.
Each surface below exists in the live HQ today — this is the product, not the roadmap.
20 named specialists across 8 departments — each with a role, a model tier, and a persistent knowledge base auto-injected on every spawn. Staff dossiers show their commits, activity, and knowledge docs. Feed any staffer new context straight from the dashboard.
One hive runs multiple companies in parallel. The Businesses view shows KPIs, open / review / shipped ticket counts, and per-project model spend with real attribution from the usage ledger. The orchestrator routes work to the right staff for the right product automatically.
Every proposal from the AI team surfaces for your review. Low impact: one confirm. High impact: your 6-digit authenticator code. Autonomy dials — provider routing, hive size, worker ceiling, spend budget — are editable from the dashboard and gated by your authenticator.
Every channel the company uses — M365 email (with de-duplication), ntfy push, Teams — is tracked with live status. Send a test email or push from the dashboard (step-up gated).
Every consequential action — loop start/stop, config edits, approvals, comms sends, knowledge feeds — is gated behind step-up MFA. The HQ observes and controls; it never runs its own executor. Nothing consequential happens without your explicit approval.
When a worker finishes a task it harvests reusable skills into a shared library, auto-injected into the next relevant prompt — the team reuses instead of regenerating. Combined with elastic headcount (workers scale to the backlog, not a fixed team size), the operation stays economical.
Real named workers, live in the system now — 20 staff across 8 departments, spanning four model tiers (Opus for the orchestrator, Sonnet for specialist engineers, Haiku for triage, GPT-4o for growth, finance and people roles).
The founder gets the whole command centre. Clients get a window into their project — without the keys.
Run the operation from a single HQ — observe everything, gate anything.
Every business on Nexus can grant its clients a scoped portal — instantly, from the founder's HQ.
Nexus is in early access. If you're a founder who wants a structured AI team building your product — with you in control of everything consequential — reach out.
No pricing yet, no sales funnel — a direct line to the founder while the product is in early access.