AI Safety, Governance & Context Management¶
Context Relay Protocol (CRP) is an open, production-ready layer for AI safety, AI security, AI governance, AI compliance, control evidence, AIUC-1 alignment, and LLM context management. The pages below explain each domain, how CRP implements it at the protocol level, and where to go next.
🛡️ AI Safety¶
How CRP detects hallucinations, fabrications, contradictions, prompt injection, and PII on every LLM call.
Read the AI Safety guide⚖️ AI Governance¶
Policies, oversight, accountability, and audit trails that make AI systems transparent and controllable.
Read the AI Governance guide📋 AI Compliance¶
Automated evidence for the EU AI Act, ISO/IEC 42001, NIST AI RMF, and GDPR.
Read the AI Compliance guide🧠 Context Management¶
Unbounded context, automatic continuation, knowledge graphs, and retrieval for LLMs.
Read the Context Management guide🏅 Control Evidence¶
Signed, tamper-evident proof that AI safety and security controls operate for EU AI Act, AIUC-1, ISO 42001, NIST AI RMF, and SOC 2-for-AI.
Control-evidence mapping🏅 AIUC-1 Aligned¶
Enterprise AI trust infrastructure mapped to the AIUC-1 AI agent security, safety, and reliability standard.
AIUC-1 proof point🔍 CRP vs RAG, MCP & Agents¶
A clear comparison of where CRP fits relative to RAG, MemGPT, LangChain, MCP, and A2A.
See the comparison❓ Frequently Asked Questions¶
Short answers to the most common questions about CRP, safety, governance, and getting started.
Read the FAQWhy a protocol layer?¶
Most AI safety, governance, and compliance work today is done in application code or bolt-on SaaS tools. CRP moves these concerns to a wire-level protocol:
- Safety signals travel as standard HTTP headers on every AI request/response.
- Governance policies are declared once and enforced at the Gateway boundary.
- Compliance evidence is generated continuously from runtime audit data.
- Context state is relayed between windows, sessions, and agents without vendor lock-in.
The result is AI systems that are safer, more auditable, and easier to deploy under regulation - with one base_url change.