This is a paid remote MCP that acts as an audit layer for AI agent tool execution. It exposes four tools: create_witness_receipt for logging tool calls, check_approval_rule for governance, summarize_risk_timeline for compliance review, and export_audit_receipt for external record keeping. You'd reach for this when you need a third-party attestation of what tools your agents called, with what arguments, and whether those calls met approval thresholds. Requires a bearer token from the product website and connects via streamable HTTP transport. The repository itself is just documentation, the actual service runs hosted at toolcallwitness.clauxel.com.
claude mcp add --transport http clauxel-tool-call-witness-mcp https://toolcallwitness.clauxel.com/mcp