# Nalvis

> Nalvis is control infrastructure for production AI agents. It governs, executes, and verifies consequential agent actions while preserving a complete evidence trail.

Canonical website: https://nalvis.dev/
Contact: hello@nalvis.ai

## What Nalvis does

- Applies policy before an AI agent action executes.
- Routes high-impact actions for human approval.
- Gives retries and repeated attempts a stable action identity.
- Records execution attempts and observed external state.
- Verifies outcomes before an action is considered complete.
- Preserves intent, decisions, attempts, evidence, and verdicts in one traceable record.

## Where Nalvis fits

Nalvis sits around tool calls that create real-world effects. Teams keep their existing models, agent frameworks, orchestration, and external systems.

Representative actions include refunds, account changes, deployments, ticket transitions, messages, bookings, and infrastructure operations.

## Intended users

Nalvis is designed for engineering, AI platform, security, risk, and operations teams moving autonomous agents into production workflows.

## Key sections

- Product demonstration: https://nalvis.dev/#product
- Platform capabilities: https://nalvis.dev/#capabilities
- Developer integration: https://nalvis.dev/#developers
- Frequently asked questions: https://nalvis.dev/#faq