wtf is compliance-agent?

ComplianceAgent is a command line tool that checks whether a codebase using AI follows the European Union's AI Act, a set of rules governing how AI systems can be built and deployed. Running one command scans a project in about 5 seconds and produces a report. The tool describes itself as a heuristic first pass rather than legal advice, and it does not by itself prove that a project is compliant. The scan works in steps. First it reads through project files looking for AI related code, such as imports from OpenAI or LangChain, or calls to an AI chat API, using actual code parsing rather than simple text search so a comment mentioning a provider name will not by itself trigger a finding. Next it assigns a risk level of minimal, limited, high, or unacceptable, based mainly on what the AI is being used for rather than how advanced it is. A resume screening or credit scoring tool would be treated as high risk, while a plain chatbot without a sensitive use case would usually be treated as limited risk. Then it checks specific articles of the EU AI Act, such as whether a project discloses that users are talking to AI or logs AI conversations for high risk uses. Finally, for each gap it finds, the tool points to a specific code fix a developer can copy in. The tool applies to anyone building with an AI provider such as OpenAI, Anthropic, or Google, or using a framework like LangChain, especially if that project is deployed in or serves users in the EU, since fines for noncompliance can run into the millions of euros. It is not needed for personal projects with no AI, or projects with no ties to the EU. Installation is done through uv or pipx as a standalone command line tool, or through pip inside a virtual environment. The project is released under the MIT license. The full README is longer than what was shown.