Paveo

Paveo: Security tool that enforces rules and checks tool calls for AI coding agents.

Paveo acts as a security seatbelt for AI coding agents like Claude Code, Codex, and Cursor by checking every shell command and file write before execution. It refuses dangerous operations such as rm -rf and DROP TABLE, logs decisions into a verifiable hash-chained record, and ensures nothing leaves your local machine. The software installs locally, requires no network sockets, and supports custom agent policies and MCP servers with built-in budgeting and rate limiting. It works with Python and integrates into existing developer workflows to catch risky behaviors and runaway loops.

Paveo at a glance

Best for: Software engineer
Runs on: macOS, Windows, Linux, API, CLI, self-hosted, open source
Pricing: Free for up to three agents per policy; paid audit evidence export.
Released: October 2026
Category: AI coding tool
Website: github.com/paveo-dev/paveo

Key features

Refuses destructive shell commands and file writes
Logs decisions into a hash-chained audit record
Enforces budgets and rate limits on model calls
Runs locally with no network sockets
Supports custom policies for self-built agents and MCP servers
Replays past sessions to test rules without execution

Use it for

Preventing AI agents from running destructive git or SQL commands
Limiting API spend and token usage for AI models
Auditing past AI coding sessions for security compliance
Guarding custom AI agents and Model Context Protocol servers

Alternatives to Paveo

Zenable — Enforces functional and security requirements across coding agents and pull requests.
Stop That Shit — Enforces task boundaries to stop unnecessary AI agent actions.
CodeGraph — It eliminates file crawling, reducing AI agent tool calls by 88% and cutting costs.
Pawl — Stops costly agent mistakes using fast deterministic code instead of prompt instructions.
Repo Guards — It prevents AI agents from performing destructive actions across multiple repositories.
jevgrep — Reduces coding agent costs by providing targeted, relevant repository context.
pstack — Improves AI coding agent outcomes with rigorous workflows.
Vne — Blocks agent access to secrets while letting it manage env files safely.
Codacy — Enforces unified coding standards and security policies across the software lifecycle.
Fluxix — Stops dangerous tool calls and data leaks before execution.
Adhdev — Orchestrates agents across machines with automated, test-gated git merging.
Anti Samcheonpo — Catches redundant AI tasks and blocks wasted agent spending.

All Paveo alternatives