---
title: "AIF overview"
description: "Learn what AIF protects, what it records, and where the product’s coverage starts and ends."
---

> Documentation Index
> Fetch the complete documentation index at: https://docs.harden.run/llms.txt
> Use this file to discover all available pages before exploring further.

# AIF overview

AIF evaluates covered coding-agent tool calls before execution and records the resulting verdict locally. It is designed to make an agent workflow inspectable: you can see why an action was allowed, blocked, paused, redacted, or recorded.

## What AIF does

- Evaluates covered actions before they run.
- Applies the decision vocabulary shown in the local viewer.
- Records local evidence so you can inspect an outcome after the action.
- Connects to supported agent integrations through their documented hooks, bridges, or proxies.

## What AIF does not claim

AIF coverage depends on the agent integration and the action surface it can observe. A decision is not a substitute for reviewing the coverage boundary of an integration or for the security controls already required by your organization.

Before relying on AIF in a workflow, read the agent-specific setup guidance and the [Trust and known limits](https://harden.run/trust/aif).

## Use the docs in this order

1. Check [system requirements](/getting-started/requirements).
2. [Install AIF](/getting-started/install) once.
3. Complete the post-install [quickstart](/getting-started/quickstart).
4. [Configure your agent](/integrations/configure).
5. Learn how to read [decisions](/reference/decisions) and [local history](/operations/local-history).

## Need an end-to-end setup?

If you need help choosing an integration, validating a machine, or reaching a first protected workflow, [book a call](https://calendar.app.google/GdL3BvBuagny6Kh5A).

Source: https://docs.harden.run/overview/index.mdx