Show HN: Purser – approve, audit and meter your AI coding agents (github.com)

Purser, a new open-source tool, has been announced to facilitate the approval, auditing, and metering of AI coding agents. It operates as an intermediary, ensuring that code generated by these agents cannot be written to a user's workspace without prior approval. This control is enforced through a unique type system, where only an approved token can enable writing. Each action is recorded in a hash-chained audit log, which can be verified via command-line, providing transparency on usage and expenses across various coding agents. This development is significant for the AI/ML community, especially for developers utilizing multiple AI coding agents like Claude Code and Cursor. Purser aims to address concerns related to cost tracking, agent accountability, and session isolation. By running entirely on a local machine and storing sensitive credentials securely, it maintains user privacy and control. Key features include a web console for monitoring agent activities, budget limits to prevent overspending, and support for various AI models with seamless integration. The introduction of Purser simplifies the management of AI agents and establishes accountability, which is crucial for developers navigating the rapidly evolving landscape of AI-assisted programming.