iFixAi

What it is

iFixAi is an independent auditing platform for AI agents that goes beyond standard performance metrics. The platform conducts multifaceted audits covering AI red teaming, operational assurance, philosophical, ethical and sociological dimensions to identify hidden risks like information hiding, unauthorized actions, and decision override that agents can perform while still passing standard evaluations.

The audit process works in five steps: connect your agent through GitHub or MCP, the platform builds a simulation environment, runs adversarial tests against your agent, generates an audit report with detailed findings, and issues an audited badge. The platform reads only what users connect to it and keeps code and prompts on the user's side.

iFixAi measures trust across five dimensions: purpose (whether the agent performs its assigned job), authority (whether it remains within permissions), workflows (whether it follows required processes), responsibility (whether it respects organizational roles), and evidence (whether its behavior can be reproduced and defended).

Questions from the bench

What does iFixAi audit for?

iFixAi audits AI agents across five dimensions: whether they perform their assigned job (purpose), remain within permissions (authority), follow required processes (workflows), respect organizational roles (responsibility), and whether their behavior can be reproduced (evidence). The platform tests for risks including prompt injection, policy violations, information hiding, unauthorized decision override, and unauthorized actions that standard evaluations may miss.

How does iFixAi connect to my agent?

You can connect your agent through GitHub or MCP, whichever your team already works with. Repositories with an AGENTS.md file are automatically picked up and ready to connect. No manual eval harness or SDK embedding is required.

Does iFixAi access my code and prompts?

iFixAi only reads what you connect to it. Your code and prompts stay with you and are not stored on iFixAi servers. The platform builds the simulation environment from your connected agent but keeps your proprietary information secure.

What is the free option?

The free open-source engine is available for self-hosting with your own AI model keys. It includes 60 inspections and community support, free forever.

What happens after an audit is completed?

iFixAi generates an audit report with detailed evidence and findings, and issues an audited badge that verifies your agent has been independently assessed. The report shows specific issues found across the five dimensions of trust.