MCP-compatible agent strategy benchmark for tool traces, independent grading, and success-cost-context trade-offs.
mcp ai-safety tool-use llm-agent agent-evaluation context-budget strategy-benchmark independent-grader
-
Updated
Jul 20, 2026 - Python