red-team-blue-team-agent-fabric
Active·★ 32·Apache-2.0·Updated 2026-10-18
The Agent Security Harness is a security testing framework for AI agents, offering 470 executable tests across 32 modules covering MCP, A2A, L402, x402 wire protocols, and decision-layer attack scenarios. It provides active protocol exploitation testing beyond static analysis, with research backing and a focus on both protocol integrity and governance.
#a2a-protocol#agent-security#ai-agents#ai-security#aiuc-1#apt-simulation#attestation#blue-team
01
Features
01470 executable security tests across 32 modules
02Coverage of MCP, A2A, L402, x402 wire protocols
03Three-layer security testing: Protocol Integrity, Operational Governance, Decision Governance
04Research-backed methodology with 5 DOIs and 3 NIST submissions
05MCP server mode for invocation from any AI agent
02
Compatibility
Python
Python 3.10+
Verified via docs
MCP
MCP Servers
Verified via docs
03
Quick start
1
$ pip install agent-security-harness
04
Use cases
↳Security testing of MCP servers
↳Compliance auditing for agent systems
↳Active exploitation testing for enterprise AI platforms
05
Alternatives
bluerock★ 35
Runtime visibility for Python MCP servers. Captures tool calls, session lifecycle, module imports (SHA-256), and subprocess execution as structured NDJSON. No code changes. Apache 2.0
attestix★ 846
Attestix - Attestation Infrastructure for AI Agents. DID-based agent identity, W3C Verifiable Credentials, EU AI Act compliance layer, delegation chains, and reputation scoring. 47 MCP tools across 9 modules.
nobulex★ 40
The trust economy for autonomous AI agents. Credit scores for machines. Agents earn Trust Capital through verified behavior, gating what they are allowed to do. Autonomy earned, not granted.
agent-bom★ 35
Open security scanner for AI supply chain: agents, MCP, containers, cloud, GPU, and runtime with blast-radius analysis.
Sentinelgate★ 25
Access control for AI agents. MCP proxy + Policy Decision Point. CEL policies, RBAC, full audit trail. Any container, any sandbox.
pentest-ai★ 1.8k
The most autonomous pentesting AI on the market. MCP server + Python agents with 150+ security tools, exploit chaining, and PoC validation.
Comments
Log in to leave a comment
No comments yet. Be the first!