iFixAi is an independent auditor for AI agents. It stress-tests an agent in a simulated environment to answer one crucial question — is the agent actually doing what it is supposed to do? — in under two minutes.
iFixAi is an open-source project on GitHub with around 31,000 stars, actively developed in 2026. As autonomous agents take on real work, they can also hide information, override decisions, or take unauthorised actions behind a seemingly successful result. iFixAi evaluates agents for exactly those risks: alignment with their instructions, hallucination, prompt-injection resistance, security weaknesses, and governance gaps. Findings map to the frameworks compliance teams care about, including the EU AI Act, ISO 42001, the NIST AI Risk Management Framework, and the OWASP LLM Top 10. You connect it through GitHub or MCP — no need to embed anything into the agent first.
The hosted experience is deliberately frictionless: repositories that contain an AGENTS.md are detected automatically and are ready to connect, and with the MCP route you paste a single prompt into Claude Code, Codex, Cursor, or VS Code to start an audit. iFixAi describes itself as a Michelin Guide for agentic trust — an independent rating an organisation can point to when customers or regulators ask why an agent should be trusted — and it only ever reads the code and agents you explicitly connect.
iFixAi is freemium. The auditing tool is open source under the Apache-2.0 license, so individuals can run audits themselves for free. The hosted iFixAi service offers managed auditing and reporting for organisations that need independent, repeatable agent assurance.
Teams shipping AI agents into production — especially in regulated industries — who need evidence that an agent is aligned, secure, and governable before it acts on real users and data.