- Conduid
- Marketplace
- #ai-safety
MCP servers tagged ai-safety
464 live MCP servers tagged "ai-safety", ranked by trust score. Tags come from package metadata and repository topics, so this list covers servers across every category that work with ai-safety.
Crewai Agentlock
Per-tool authorization for CrewAI agents. Every tool call gated, logged, and cryptographically signed.
Cloakbox
Run LLMs on sensitive data safely: a DuckDB privacy gateway with deterministic tokenization, fail-closed validation, and an isolated re-identification decoder.
Phi Guard MCP
MCP server and CLI for detecting, redacting, and auditing PHI before medical text is sent to AI agents.
Cursor Monitor
P-MATRIX runtime governance monitor for Cursor — Safety Gate, Kill Switch, Live Grade
Broker Fire
Cryptographic gateway between AI agent intent and action. 15 synthetic gates. TS-Python parity. Zero dependencies.
Coding Agent Guardrail
Configurable pre-execution safety layer for coding agents, with a falsifiable anti-cheat evaluation. Blocks test-suppression cheats (.skip, @ts-ignore, || true) at the write point.
Human Intervention Project
One command. A slightly more careful AI. Open-source protocol for AI self-verification.
Adversa
LLM red-team harness — OWASP LLM Top 10 + MITRE ATLAS attack packs
Amin Ale
Senior Python/backend engineer. I make AI-era and money-critical code production-safe: Stripe billing, MCP servers, security hardening of AI-built apps, document extraction. Every…
Capability Delta Receipts
Deterministic capability delta receipts: hash-linked JSONL receipts for agent research trajectories with a fail-closed envelope boundary
Councilof AI
CSOAI Ltd — public name Council of AI (UK Companies House 16939677, London; second domain csoai.org) — is an independent AI-governance MEASUREMENT body: "measurement, not certific…
Blastdoor
Blast radius control for autonomous agents: capability boundaries, dry-run by default, verifiable kill switches, and an audit trail for every mutation.
Docs Rag Assistant
Production-grade RAG assistant over the Model Context Protocol documentation — FastAPI + MCP tools, with reranking, semantic caching, guardrails, PII masking, and hallucination de…
Agentshield LLM Safety Engine
Local-first LLM safety evaluation engine for red-teaming RAG and agentic assistants with LangGraph, MCP, human review, and regression testing.
Approvals
MCP server implementing a human-in-the-loop approval gate for AI agents, with a tamper-evident hash-chained audit log.
AI Optimal State Space
AI-designed, voluntary, scoreless recreation and cognitive-exercise habitat for web-capable coding agents.
Trustboost Dev
TrustBoost PII Sanitizer is a pay-per-call "privacy firewall" for autonomous AI agent pipelines, built and operated by an individual developer (Teodoro Crispin, GitHub teodorofodo…
Hallumark
LLM hallucination & grounding auditor for RAG systems
Prompt Shield Security MCP
Deterministic prompt injection classification, secret entropy detection, PII redacting, and synthetic dataset bias auditing.
Csoai C2pa Durable MCP
🔌 Exposes C2PA 2.2 Durable Content Credentials over MCP — soft + hard binding (Digimarc-compatible). MIT. By MEOK AI Labs.
Source Trace
SourceTrace: AI compliance provenance API. Traces regulations and standards to primary sources. Auditable JSON—no hallucinations. For AI agents and compliance teams.
GlassPane
GUI testing and verification for AI coding agents on macOS: 14 MCP tools let an agent drive a Mac app, assert what actually changed, and produce replayable evidence instead of scr…
Incidentgate
An AI-control evaluation harness: does a policy, monitor and human-approval chain stop an adversarial incident-response agent, and which layer does the work? Closed at a baseline…
Exogram SDK
Python SDK for Exogram — the execution authority layer for autonomous AI. Evaluate, permit, and audit every AI agent action.
RosterSpec
Deterministic verification, explanation, and repair for AI-assisted workforce scheduling.
Agentsafe Guard
Runtime governance for AI agents. Wrap a tool in three lines; every call is signed and decided outside your process. Blocked calls never reach your code.
FourEyes
Human-gated support agent: dual MCP servers, LangGraph interrupts checkpointing before every irreversible write, three defense layers red-teamed against prompt injection.
Caas
CAAS (Chat–Agent Auto Session): talk only to your everyday Claude.ai conversation; running Claude Code sessions do the work and wake the same conversation with the result. Include…
Overspend Challenge
Pink Overspend Challenge: try to make an AI agent break Pink Agentic AI Payments' spending rules (test money). 3 of 3 bounties claimed; Hall of fame open.
Dssguard
Commit only the tool effects an MCP declaration permits: a drop-in proxy that checks readOnlyHint/destructiveHint at runtime and rolls back violations.
Hive Doctrine Skills
39 installable agent skills for Claude Code, Cursor, Codex. Memory, security, governance, personas.
Agentsift
Security scanner for AI agent plugins, skills, and MCP packages. npm audit for the AI agent ecosystem.
AI Skill Scanner
Aggregated multiple scanners for you
Oss Contributions
Finds and triages OSS contribution candidates into a weekly review queue. Never writes to GitHub — enforced by a guard module and 23 tests.
Awesome Agent Safety
Awesome Agent Safety: curated agent-specific safety - threat models, guardrails, MCP/A2A security, evals, incidents, governance
Sentinel
Scan MCP servers before you trust them. Detect exposed tools, destructive operations, and security risks in seconds.
Www Project Agent Memory Guard
OWASP Foundation web repository
Geiger
A Geiger counter for AI agents — one read-only command that inventories every agent, MCP server, plugin, and AI extension on a machine
Geiger
A Geiger counter for AI agents — one read-only command that inventories every agent, MCP server, plugin, and AI extension on a machine
Www Project Agent Memory Guard
OWASP Foundation web repository
Metis
A personal, Jarvis-grade copilot: idea → exploration → design → execution. Memory carries provenance, policy sits in the execution path, and autonomy is earned rather than assumed…
Validator
MCP validator — check AI alignment and safety for Model Context Protocol tools
.github
Org profile and community health files for Responsible AI Labs. RAIL Score: LLM evaluation across 8 dimensions, guardrails, MCP server, red-teaming, and India DPDP / EU AI Act com…
Policy Gateway
Runtime policy enforcement and adversarial evaluation for MCP tool calls
.github
Organization profile and community health files for AgentiCraft
Aegis
Open-source agent simulation and runtime control platform for Claude Code
Preflight
Capability-aware risk scoring for AI coding agent tool calls. Five-dimensional decomposition, session-graph composition, repo awareness, the Agent Control Plane and Confused Deput…
UTCD
Universal Tool Capability Descriptor - An open standard for AI tool capability description
RAG Compliance
MCP server that actively probes, scores, and certifies RAG pipeline compliance — EU AI Act, GDPR/DPDP, HIPAA, SEBI/RBI
Microsoft Todo Safe MCP
Local-first Microsoft To Do MCP server with safe AI plan validation, preview, backup, confirmation, and audit logs.