This is a hosted remote MCP that generates permission boundary documentation and governance receipts for ChatGPT agents. It exposes four tools: map_permissions, flag_sensitive_fields, summarize_boundary, and export_remediation_plan. You'd reach for it when you need audit trails showing what data an agent touched, which fields are sensitive, and what remediation steps to take. It's a paid service requiring a bearer token from agentdataboundary.clauxel.com, runs over streamable HTTP, and returns structured evidence you can hand to compliance or security teams. Think of it as a black box recorder for agent data access.
claude mcp add --transport http clauxel-agent-data-boundary-mcp https://agentdataboundary.clauxel.com/mcp