#ai-safety

The Execution Security Layer for the Agentic Era. Providing deterministic "Sudo" governance and audit logs for autonomous AI agents.

ThumbGate Pre-Action Checks derive rules from repeated failures, flag risky tool calls, hard-block detected secret leaks, and block matches in strict mode.

Grade every agent, scraper and model in a claim's chain — provenance, trust scoring and audit evidence for LLM pipelines

Zero-trust governance layer for Claude Code. Policy-enforced MCP gateway with autonomy controls, middleware chain, audit log, HALT kill-switch, and prompt-injection defense.

MCP servers expose tools with no information about what they actually do at runtime. mcpsafetywarden sits between your agent and any MCP server, profiling tool behavior, blocking destructive calls, and running active security audits before you trust them in a workflow.

Make AI coding agents safe to scale autonomously: assign work, cap spend, enforce policy, verify output, roll back failures, learn from loops, and prove ROI across every repo.

License Compliance Checker — Multi-ecosystem license + AI model scanner for EU AI Act Article 53 GPAI compliance. SBOM, SARIF, training-data risk. Apache 2.0.

AuspexAI volunteer worker for donating compute to AI research

Offline cryptographic receipts for AI evaluation results — Ed25519 + RFC 6962 Merkle + optional SD-JWT. Integrity, not truth

A lightweight, highly secure AI API Gateway/Proxy written in Go. Acts as transparent middleware between local AI coding clients (OpenCode/Pi/Cursor) and upstream LLM providers (Gemini, DeepSeek, Zhipu z.ai).

Deterministic governance for Claude and MCP tool calls. Pin approved capabilities, detect drift, and refuse unsafe or unapproved actions before execution. No model in the decision path.

Cognitive Browser: The browser automation that thinks. Constitutional safety • Persona UX testing • Natural language interface • Self-healing selectors • Built for AI agents

SDK for authoring research tenants on AuspexAI

pytest for AI agents - Autonomous red-teaming, behavioral monitoring & security testing for LLM agents

Verification for the agent era. Your coding agent's PR summary cannot lie about its diff - one CI line. Plus the research: the first map of which AI minds can read each other. Every claim machine-verified against committed receipts, negatives included. pip install styxx

The WHY layer for AI agents: six constitutional gates + 12 hard constraints, plus cross-session risk composition — catches agents that pass every individual gate but accumulate risk across a sequence (stateless engines can't). pip install constitutional-agent

Per-operation approval and audit logging for secret access and git commit signing on Linux