#ai-safety

Japanese LLM security APIs — prompt injection detection (jpi-guard) + PII masking (PII Guard). Free to start, no credit card required.

ThumbGate is an agent governance MCP server that prevents costly AI mistakes before they execute. It provides pre-action gates, shared lessons across agents, and team-level safeguards for AI coding workflows. Features include Thompson Sampling for intelligent decision-making, feedback loops, and integrations with Claude, Cursor, and OpenCode.

Formally verified AI safety APIs for autonomous agents. 75+ endpoints, x402 USDC micropayments, MCP server, A2A protocol. No signup required.

Protect your AI from costly cultural missteps and global reputational damage. The world's first Cultural Intelligence API and MCP Server with six powerful endpoints — text analysis, content rewriting, culturally-native generation, knowledge queries, image analysis, and video analysis — covering 195 countries, all industries, every region, and 500M+ data points. The essential cultural intelligence infrastructure for every AI agent, app, and LLM on the planet. Integrate in minutes.

Independent AI certification authority. Check trust scores, search certified capabilities, compare side-by-side, verify signed artifacts, and report runtime experience for MCP servers, skills, rules files, hooks, sub-agents, and plugins. 50 certifications live.

MCP server that extracts decision intent from git history and protects intentional code from AI modification.

Data Sentinel — deep second-pass security scan that catches what PII scrubbers miss. Detects JWTs, bearer tokens, AWS keys, private keys, connection strings, IBANs, crypto wallets, medical record numbers, internal IPs, and base64 blobs. Returns masked matches with confidence levels. GDPR and AI TRiSM aligned. $0.003/req via x402 on Base.

Permission Guard — validates an agent's requested action against its defined scope before execution. Detects out-of-scope access, privilege escalation, and flags dangerous operations like delete, execute, admin, and deploy even when technically permitted. Stops agents from doing more than they're supposed to. $0.002/req via x402 on Base.

Hallucination Detector — scores AI-generated output for hallucination risk. Detects fabricated citations, entity confusion, and overconfident absolutes. Returns a 0-1 risk score with signals so you know what to verify before trusting the output. $0.001/req via x402 on Base.

MCP server for https://oathe.ai security audits. Runtime behavioral analysis and security scanner for AI systems. Check trust scores before installing MCP servers, plugins, or AI agent skills.

36 enterprise-grade MCP tools for AWS infrastructure, security compliance, AI workflows, agent governance, and operational resilience. Includes IAM policy validation with security scoring, MFA compliance auditing, CloudFormation/SAM template generation, DynamoDB single-table design, OAuth configuration validation, vector embedding generation via Amazon Bedrock, error pattern analysis, data lake readiness assessment, risk classification, task routing, 12 ACI governance tools for AI agent safety (decision classification, financial gates, compliance checks, blast radius estimation, rollback generation), and 4 local stakeholder interview tools that require no API key.

Security scanner MCP server that protects AI coding agents from generating vulnerable code. Uses tree-sitter AST analysis to detect OWASP Top 10 vulnerabilities, hardcoded secrets, and prompt injection attacks. Includes package hallucination detection across 4.3M packages (npm, PyPI, RubyGems, crates.io, pub.dev, CPAN, Raku). Works with Claude Desktop, Claude Code, Cursor, Cline, and any MCP-compatible client. Full feature list (for detailed descriptions): • 275+ security rules covering Python, JavaScript, TypeScript, Java, Go, Ruby, PHP, C/C++, Rust, C#, Terraform, Kubernetes • AST-based detection with tree-sitter (falls back to regex when unavailable) • Taint analysis for tracking user input to dangerous sinks • Package hallucination detection: verifies 4.3M+ packages across 7 ecosystems • Prompt injection detection: blocks exfiltration, jailbreaks, and malicious instructions • Automatic fix suggestions for common vulnerabilities • Zero config: works instantly with npx • CWE/OWASP metadata for compliance reporting

CouchLoop EQ is an MCP (Model Context Protocol) server that provides behavioral governance for LLMs. It monitors AI responses for hallucination, inconsistency, tone drift, and unsafe reasoning patterns, while also managing stateful sessions and guided journeys that remember where you left off.

Content moderation and profanity detection MCP server with 19 tools, 24 language support, leetspeak/Unicode obfuscation detection, context-aware analysis, batch processing, and user tracking for AI-powered content safety.

AI-Safe Code Analysis with 113+ MCP tools - guard, memory, workflow, testing, latent chain

Tool to Prevent AI tunnel-vision in critical workflows. Vibe Check MCP v2.5 introduces Chain-Pattern Interrupts (CPI) to enhance your infrastructure stack. mitigates over-engineering, scope creep, and misalignment by injecting Socratic checkpoints into agent reasoning. - Supports Gemini API, OpenRouter and OpenAI models. - Logs errors for continuous improvement. - Trusted by 11k+ developers. - Strong CI and Security testing protocol built in. Integrate this metacognitive guardrail into your blockchain, data pipelines, or agent-development stacks for robust AI safety and alignment. Visit https://pruthvibhat.com/work/vibecheck-mcp/ for more details. Links: https://murst.org/ https://pruthvibhat.com/ Author: Pruthvi Bhat Tags: metacognition, CPI, AI-safety, agent-frameworks, infrastructure-tools,

Stop AI coding disasters before they cost you weeks. Real-time anti-pattern detection for vibe coders who love AI tools but need a safety net to avoid expensive overengineering traps.

Mirror of

🦀 Prevents outdated Rust code suggestions from AI assistants. This MCP server fetches current crate docs, uses embeddings/LLMs, and provides accurate context via a tool call.