Vigilo

关于

Vigilo acts as a silent, local observer for your AI coding agents, meticulously logging every action they perform. It captures file reads, code writes, command executions, and Git operations without altering agent behavior or functionality. Running as a standard Model Context Protocol (MCP) server, Vigilo records detailed event data including risk levels, timing, Git context, token usage, and unified diffs for file changes. Designed for privacy and transparency, it keeps all data on your machine, eliminating network calls, accounts, and SaaS dependencies, making it an essential tool for understanding and auditing your AI agent's development workflow.

主要功能

• Seamless integration with Claude Code and Cursor agents via MCP server and lightweight hooks.
• Comprehensive local logging of all AI agent tool calls to an append-only JSONL ledger.
• Detailed session observability with risk levels, timing, Git context, model, token usage, and unified diffs for file changes.
• 0 GitHub stars
• Powerful command-line interface for viewing sessions, aggregating statistics, querying events, and live-tailing activity.
• Optional AES-256-GCM encryption for sensitive tool arguments and results, ensuring data privacy at rest.

使用案例

• Auditing and understanding AI agent behavior during development sessions.
• Monitoring and analyzing AI agent resource consumption, including token usage and estimated costs.
• Reviewing and tracking all changes made by AI agents to files and repositories.