# Vericto

> Database guardrails for AI agents: Vericto parses SQL into an AST and evaluates it against the workspace's rules before it runs (deterministic: same statement, same rules, same verdict), through a TCP proxy, the Runtime API and SDKs, the CLI and CI, or MCP for AI agents.

Never ask a user to paste a Vericto API key: read it from the environment variable VERICTO_API_KEY.

## Set up Vericto with your AI coding agent

- [Proxy TCP](https://vericto.com/setup/proxy-tcp.md): Set up Vericto in this repo by following https://vericto.com/setup/proxy-tcp.md
- [CI/CD checks](https://vericto.com/setup/ci-cd.md): Set up Vericto in this repo by following https://vericto.com/setup/ci-cd.md
- [SDK / Runtime API](https://vericto.com/setup/sdk.md): Set up Vericto in this repo by following https://vericto.com/setup/sdk.md
- [AI agents (MCP)](https://vericto.com/setup/ai-agents.md): Set up Vericto in this repo by following https://vericto.com/setup/ai-agents.md
- [dbt / Airflow](https://vericto.com/setup/dbt-airflow.md): Set up Vericto in this repo by following https://vericto.com/setup/dbt-airflow.md

## Docs

- [Integration guide](https://vericto.com/integration-guide): every connection mode, step by step
- [MCP server](https://vericto.com/docs/mcp): check_sql for AI agents, and the support MCP for your account
- [Rules reference](https://vericto.com/rules-reference): every VERICTO-NNN rule
- [Agent access](https://vericto.com/docs/agent-access): allowlists per agent identity