failproofai

Observability and enforcement for AI agent harnesses. Capture every run and runtime reliability with policy enforcement.

What is failproofai?

Observability and enforcement for AI agent harnesses. Capture every run and runtime reliability with policy enforcement.

Overview of failproofai

What it is

FailproofAI is an observability and enforcement tool for AI agent harnesses. It hooks into 12 agent harnesses—including coding CLIs like Claude Code and Codex, chat gateways like Hermes, and self-hosted assistants like OpenClaw—to capture every run and block dangerous tool calls before they execute. It offers 40 built-in policies, runs locally, and claims zero latency.

Key features

  • Hooks into 12 agent harnesses across two classes: ten coding CLIs and two chat/assistant gateways (Hermes, OpenClaw).
  • Captures every run and enforces policies that can block tool calls before they execute on all 12 harnesses; turn-end gates are supported on eight.
  • Provides 40 built-in policies, including blocking reads of secret files, privilege escalation, recursive file deletion, and unreviewed infrastructure changes.
  • Allows custom policies written in JavaScript, placed in .failproofai/policies/.
  • Offers a local dashboard on localhost:8020 with no account or signup required, showing session history, model calls, tool calls, and hook decisions.
  • Includes a hosted observability service for teams, with features like execution graphs, latency tracking, cost tracking, error tracking, SQL over traces, and alerts.

Use cases

FailproofAI is designed for developers and teams running AI agents in coding CLIs, chat gateways, or self-hosted assistants. It can be used to enforce safety policies, audit agent behavior, and prevent dangerous actions such as file deletion, privilege escalation, or unreviewed infrastructure changes. The local dashboard is suitable for individual use, while the hosted service supports fleet-wide observability and compliance.

Getting started

shnpm install -g failproofaifailproofai config # wire up your agents and the daemonfailproofai policies add FailproofAI/policies # choose what to enforcefailproofai # dashboard on localhost:8020

License and pricing

The license is MIT with Commons Clause—free for internal and personal use; commercial resale of failproofai itself requires a separate agreement. The pricing type is open source. The hosted observability service is available with an Enterprise plan for self-hosting in your own cluster.

Based on the project's public README and repository metadata.

Add this badge to your README

Featured on YunoDB Badge styles & help

Maintain this project? Show that it is featured on YunoDB.

Markdown[![Featured on YunoDB](https://yunodb.com/badge/failproofaifailproofai.svg)](https://yunodb.com/tools/failproofaifailproofai?utm_source=github&utm_medium=badge)
HTML<a href="https://yunodb.com/tools/failproofaifailproofai?utm_source=github&utm_medium=badge"><img src="https://yunodb.com/badge/failproofaifailproofai.svg" alt="Featured on YunoDB" height="20"></a>