Agent Evidence is an agent-first Model Context Protocol (MCP) server designed to provide robust visibility and auditability for automated tasks. It allows AI agents to automatically capture visual evidence, such as screenshots and screen recordings, at critical junctures, especially during complex desktop or browser operations. Beyond mere capture, it enables the annotation of artifacts with notes or OCR text, and the redaction of sensitive information for secure sharing. The tool organizes all output into a cohesive package including a summary, timeline, and all collected artifacts, ensuring a clear, reviewable record for debugging, QA, or client handoff.
主要功能
01Organize all evidence into a reviewable output including a summary and timeline
02Integrate seamlessly with AI agents to automate evidence collection
03Capture milestone screenshots and short screen recordings during agent tasks
04Attach notes or OCR text to captured visual artifacts
051 GitHub stars
06Redact sensitive areas from screenshots to generate shareable copies
