AI_glue – drop-in audit and governance for OpenAI and Anthropic apps (github.com)

AI_glue has been announced as a new tool designed to enhance observability and governance for applications utilizing OpenAI and Anthropic models. This drop-in solution allows every API call to be logged, including details such as provider, model, tokens consumed, associated costs, and latency. Key features include visibility into spending by project and team, detection of personally identifiable information (PII) during individual calls, enforcement of governance rules, and a unified dashboard for monitoring across various environments. The implementation requires only a single environment variable change, enabling organizations to integrate governance into existing applications without needing extensive code rewrites. The significance of AI_glue lies in its capacity to streamline compliance and oversight for AI deployments, addressing the growing need for transparency and accountability in AI application usage. By acting as a transparent proxy between applications and model providers, it facilitates robust auditing and policy enforcement while allowing teams to manage operational costs effectively. It supports multi-instance aggregation for development, staging, and production environments, making it easier for organizations to oversee AI usage across all stages of the application lifecycle, ultimately assisting in data security and budget management.