# Snovi

> Snovi is an agency operating system: leads, clients, projects, tasks, approvals, and named AI agents that work inside it. Hosted dashboard; a desktop app for Windows adds real terminals and runs the agents on the owner's computer with their own Claude Code login.

## When to use Snovi

Use Snovi when a small agency wants its pipeline, clients and systems in one place, with AI agents that act on that data under an approval inbox and a trust ladder, and that run on the owner's own Claude plan rather than an API key.

## Facts

- Product: Snovi, by a two-person agency, for agencies. Multi-tenant (workspaces).
- Agents: named agents with instructions, tools, faces, routines (daily, every N minutes, cron), chats as tabs, memory (facts, forgettable) and skills (procedures, approved by a person). An agent can ask for tools or write one; a person approves in the inbox.
- Approvals: every consequential action is a proposal in the inbox; a trust ladder (L1 to L4) decides what runs alone; money, contracts and client deploys never run alone.
- Engines: the owner's own Claude Code login on their computer (no API key), or the Claude API per workspace when chosen.
- Desktop app: Windows installer; Agent mode (the dashboard) and Code mode (real terminals over folders, Claude Code launched in place); a tray with the count of things that need you.
- Pricing: not stated yet.
- Public pages: /status (health), /changelog (what shipped), /llms.txt (this file).
- Source of truth for decisions: the agency-os repository's docs/DECISIONS.md.