# Vivari

> Vivari is the AI agent workspace: the environment where fleets of AI agents are put to work and kept accountable. Agents arrive knowing your world, work as a team in rooms that carry context, roles, and review, and earn trust one reviewed change at a time. A human can watch the whole fleet in one glance, step inside any live session, and take over at any moment.

Vivari runs the agents you already use, with cloud models or your own: Claude Code, Codex CLI, and Cursor CLI today, plus local and custom OpenAI-compatible endpoints through the same harnesses. Vivari Guard, the deterministic review engine, scores every risky change against a deep map of your whole codebase, with no LLM in the call; verdicts are PASS, WARN, or BLOCK in pre-commit hooks and CI. Everything an agent does lands in a replayable audit trail. Vivari never uploads or mirrors your repositories and never trains on your code; Guard's analysis runs locally with no network call. Founded by Wael Masri (https://waelmas.com). Curated early-access cohorts open fall 2026; pricing is announced at launch. Contact: hello@vivari.ai

## Pages

- [Home](https://vivari.ai/): what Vivari is, the fleet-to-agent descent, your first Monday, what changes by Friday, early access
- [What is AI agent management?](https://vivari.ai/what-is-ai-agent-management/): the guide - definition, how it differs from orchestration, AgentOps, and governance, the seven disciplines, questions your stack cannot answer today
- [Method](https://vivari.ai/method/): nine principles for running agent fleets accountably
- [Manifesto](https://vivari.ai/manifesto/): A Brilliant Stranger in Your House - why the missing layer is management
- [Vivari Guard](https://vivari.ai/guard/): deterministic change-safety for AI-authored code - the deep codebase map, weighted signals, PASS/WARN/BLOCK verdicts, local-first
- [Security](https://vivari.ai/security/): local-first architecture, scoped per-agent permissions, deterministic review, audit trail with SIEM export
- [The First Week](https://vivari.ai/stranger-test/): a playable test - eight decisions across an agent's first week, scored 0 to 16, free, no email required
- [Glossary](https://vivari.ai/glossary/): dated definitions - agent memory, permissions, audit trails, governance, observability, the inner and outer loop, and more
- [Open](https://vivari.ai/open/): the receipts - pages on this site shipped by a fleet of agents managed in Vivari, with the trust decisions and review verdicts behind them
- [Pricing](https://vivari.ai/pricing/): announced at launch; early cohorts lock founding terms
- [Founder](https://vivari.ai/founder/): why Vivari Guard is deterministic and local-first, from the founder

## Optional

- [Privacy](https://vivari.ai/privacy/): what is collected and why, GDPR and CCPA
- [Cookies](https://vivari.ai/cookies/): no tracking cookies today; what changes if that ever changes