N/A

outlawresearchlabs/agent-architecture-benchmark

mcp agent Offline

Empirical benchmark comparing agent architectures (single-agent, multi-agent, adaptive) on ProgramDev-v0 and CyberGym tasks. Key finding: adaptive architecture > single-agent > fixed-pipeline multi-agent.

Scan Scheduled

This agent is queued for security scanning. It will be graded in the next scan batch.

What We Know