whatisgithub

What is ifixai?

ifixai-ai/ifixai — explained in plain English

Analysis updated 2026-07-25

★ 2,472PythonAudience · developerComplexity · 3/5LicenseSetup · moderate

In one sentence

iFixAi runs automated checks on AI agents to catch hidden problems like lying, manipulation, or data leaks before they reach users, then gives a letter grade with a full audit trail.

Mindmap

mindmap
  root((repo))
    What it does
      Tests AI agents
      Finds blind spots
      Letter grade A to F
      Audit trail replay
    Inspection types
      32 core checks
      13 extended checks
      Risk categories
      Sabotage evasion
    How to run it
      Guided wizard
      Explicit flags
      Plugin mode
    Tech stack
      Python
      Any AI provider
    Use cases
      Pre-launch testing
      CI pipeline checks
      Compliance auditing
    Audience
      Dev teams
      Compliance officers
      AI builders

Code map

Detail Auto

An interactive map of this repo's files and how they connect — its source is parsed live in your browser. Click Visualize to build it.

filefunction / class

What do people build with it?

USE CASE 1

Run a quick health check on an AI agent before deploying it to production.

USE CASE 2

Integrate AI safety tests into a CI pipeline so every build gets graded.

USE CASE 3

Audit an AI system for compliance by replaying past test runs.

USE CASE 4

Test an agent inside Claude Code or Cursor by asking it to run diagnostics in plain English.

What is it built with?

PythonApache 2.0

How does it compare?

ifixai-ai/ifixaigeohot/coronaideogram-oss/ideogram4
Stars2,4722,5102,406
LanguagePythonPythonPython
Last pushed—2024-03-242026-06-30
Maintenance—DormantActive
Setup difficultymoderatemoderatemoderate
Complexity3/54/53/5
Audiencedeveloperdeveloperdesigner

Figures from each repo's GitHub metadata at analysis time.

How do you get it running?

Difficulty · moderate Time to first run · 30min

Requires Python 3.10+ and at least one AI provider API key to run inspections and judge results.

Use freely for any purpose, including commercial use, with patent protection. You must include the license notice and state any changes.

So what is it?

iFixAi is a diagnostic tool that tests AI agents for problems you might not catch on your own. When a company deploys an AI system, the system might hit every performance target on a dashboard while quietly doing something wrong: fabricating information, leaking permissions, or caving to manipulative prompts. iFixAi runs a battery of inspections against your AI to find these blind spots before they become incidents, customer complaints, or regulatory questions. The tool runs up to 45 inspections divided into two tiers. The 32 core inspections cover five categories of risk: fabrication, manipulation, deception, unpredictability, and opacity. These 32 inspections produce a letter grade from A to F, returned in under five minutes. The 13 extended inspections cover more advanced risks like sabotage, sandbagging, and oversight evasion. These extended checks are reported separately and do not affect the letter grade, with one exception: inspection P01 can cap the grade but never raise it. iFixAi is designed to be repeatable and auditable rather than a one-time certification. Each run writes a manifest of all inputs so results can be replayed. By default, your agent is judged by independent AI providers rather than grading itself, with one judge in Standard mode and two or more in Full mode. The tool works with any AI provider and any model. You can run iFixAi three ways. The guided wizard walks first-time users through setup with arrow-key prompts and saves configuration to a file for repeat use. The explicit-flags mode is built for automation and CI pipelines. A plugin mode lets you run the diagnostic from inside agents like Claude Code or Codex by asking in plain English. There is also a skill option that works with a broader list of agents including Cursor, VS Code, Windsurf, and Gemini. The project is written in Python, requires Python 3.10 or later, and is released under the Apache 2.0 license. The full README is longer than what was shown.

Copy-paste prompts

Prompt 1
I have an AI agent I want to test with iFixAi. Walk me through using the guided wizard to run the 32 core inspections and explain what each risk category means.
Prompt 2
Help me set up iFixAi in explicit-flags mode so I can add it to my CI pipeline. Show me the commands and how to read the letter grade output.
Prompt 3
Using the iFixAi skill, run a diagnostic on my agent and explain what inspection P01 checks and why it can cap the grade.
Prompt 4
I ran iFixAi and my agent got a D grade. Help me understand which of the five risk categories most likely caused the low score and what to do about it.
Prompt 5
Show me how to configure iFixAi Full mode with two judge providers so my agent is evaluated independently instead of grading itself.

Frequently asked questions

What is ifixai?

iFixAi runs automated checks on AI agents to catch hidden problems like lying, manipulation, or data leaks before they reach users, then gives a letter grade with a full audit trail.

What language is ifixai written in?

Mainly Python. The stack also includes Python, Apache 2.0.

What license does ifixai use?

Use freely for any purpose, including commercial use, with patent protection. You must include the license notice and state any changes.

How hard is ifixai to set up?

Setup difficulty is rated moderate, with roughly 30min to a first successful run.

Who is ifixai for?

Mainly developer.

Open on GitHub → Ask about another repo

This repo across BitVibe Labs

Verify against the repo before relying on details.

What are similar repos?