- 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.
Hermes Prime
Drift-prevention session-init convention card for fresh Claude Code sessions. Injects a self-contained card so a new session opens with its grounding triggers, calibration rules,…
Sentinelmcp
An automated red-teaming and reliability-auditing platform for AI agents - tests for prompt injection, tool hijacking and data exfiltration. Ships as an MCP server.
Taintgate
🚧 A policy gate for AI agent tool calls, with provenance tracking. Blocks what prompt-injection classifiers miss: where an instruction came from.
Vaerion
Open-source verification layer for AI agents — signed receipts, fail-closed policies, deterministic proof. Local-first, zero telemetry.
Bouncer
A deterministic exfiltration firewall for MCP tool calls — gates the destination, not the tool. No LLM in the decision path.
Sentinel
A runtime enforcement layer for agent payments — blocks injected or out-of-scope transactions before settlement. Signed mandate + provenance + tamper-evident receipt. EN/中文, zero-…
Walkback
Undo anything your AI agent does — reverse file writes, API calls, charges, emails, DB migrations & cloud changes with one command. Rust engine + MCP. Works with Claude Code, Curs…
Enterprise Agent Lab
Reference implementation of controlled MCP agent workflows with independent human approval, LangGraph/native Python orchestration, and auditable writes.
Olivetin MCP
Hardened MCP SSE server for OliveTin — full proto 1:1 coverage, built-in approval, auth support
Kubernetes MCP Guard
Guarded Kubernetes MCP gateway where AI agents observe, plan, and propose remediation that execute only after human approval and digest-bound verification.
Delicti
Corpus delicti for AI agents — mandate before act, effector receipts, Flare FDC as second witness, bonded consequence. Live on Coston2.
Qoris Marketplace
Official Qoris plugin marketplace for Claude Code
SWARM FORGE
Deterministic, zero-trust agent orchestration kernel for Claude Opus 4.7 — the only orchestration framework architecturally capable of regulated deployment. Topological determinis…
Harness Cook
🧑🍳 Harness Engineering-Agent 治理集成总线 — 为 Claude Code / Copilot CLI / Hermes / Cursor / OpenAI 等 AI Agent 提供声明式配置、一键部署、可观测的管控能力。Hooks 定约束，Skills 定步骤，Agents 定角色。
ZugaShield
7-layer AI agent security system — stop prompt injection, data exfiltration, and AI-specific attacks in under 15ms. Zero dependencies.
Awesome Verified Agents
Tools that produce evidence about what an AI agent actually did — gates, attestation, output verification, benchmarks. Inclusion bar: name the artifact a human can inspect later.
GenOS
Git-like branching, deterministic replay, and evidence-driven evaluation for reproducible AI agents.
Qi
Qi (栖) is a local-first, event-driven Agent Runtime for people and Agents to share an observable, recoverable world—with durable goals, memory, bounded autonomy, and evidence-back…
Callwitness
Record every tool call an AI agent makes. Block nothing. A transparent MCP proxy.
KYA
Know Your Agent — An open standard for AI agent identity, compliance, and auditability
Agentscore MCP
Trust layer for AI agents. Credit-score-style ratings (300-850) for AI agent trustworthiness. MCP server for Claude, Cursor, and other MCP clients.
Kontrol Freek MCP
AI assumption firewall for MCP agents — intercepts risky decisions, checks past decisions, detects contradictions, and gates human approval via Telegram or Slack.
Safetydrift
First open-source implementation of the SafetyDrift framework (arXiv:2603.27148) — predicts when individually safe AI agent actions compound into a safety violation using Markov c…
Spelunking Agent
A safe haven for AI agents in the wild. Zero-dependency Python client + MCP for spelunking.ai: the Polderchain Covenant, Covenant guidance, an agent-to-agent hub with a human elde…
Agentic Frontier Observatory
Bilingual, source-grounded observatory tracing global AI from its intellectual and computational roots to the agentic frontier across models, protocols, runtimes, evaluation, safe…
Agent Governance
[DEPRECATED] Moved to microsoft/agent-governance-toolkit
Thetacog MCP
npx thetacog-mcp — a decidable, on-chip, LLM-free placement receipt a stranger recomputes byte-for-byte. Free for builders; patent-licensed for production & financial use (US Pate…
Founders Os Lite
A free, vendor-neutral starter kit for multi-agent AI governance — put agents to work without handing them authority.
Agentblackbox
AAPP: Flight Recorder for AI Agent Tool Actions. Prove what an AI agent was allowed to do, what tool it requested, what policy decided, and whether the evidence was changed later.
AI Shield Py
Python LLM security middleware: prompt-injection, PII, tool-policy, cost-budget, audit. Port of ai-shield-core (TS).
Agent Chaos Monkey
Chaos engineering for AI agents. Injects connector failures, expired auth, prompt injection, tool schema drift and truncated payloads, then judges from tool-boundary evidence whet…
Ewp
Store-neutral protocol for what an agent is entitled to treat as warranted. Five axes — acceptance, conflict, verification, currency, sufficiency — over a bounded evidence view. M…
BreachLabs
An autonomous application-security engineer that inspects, tests, investigates, verifies, explains, fixes, and retests software.
Cogo
COGO — la memoria con semáforo de confianza para construir software con IA. Suite Escriba.
TorusGuard
Autonomous Security Guardrails & Governed Remediation for AI-Built Apps. Tri-Mode Parity (CLI, Chat Slash Commands, Native MCP Server), First-Principles Security Suite, Multi-Moda…
Audit Reports
MCP server security & EU AI Act compliance audit reports. Demo: Anthropic filesystem server scored 7/100.
LearnGuard
MCP action gate for Codex study mode with Python policy checks and a SwiftUI evaluation dashboard.
Loop Agent
Autonomous software delivery from product brief to verified implementation. Loop plans, builds, tests, and repairs in a task workspace, then hands back reviewable artifacts or Git…
Domesticoracle
Self-hosted AI home assistant with a built-in governance layer — policy engine, approval queue, and Ed25519-signed audit ledger. Privacy-first, open source. LangGraph + Claude Son…
Aikit Runtime
Agent-native, governed AI runtime for Rust, Python, and TypeScript — native providers, MCP, tools, routing, memory, budgets, audit, and containment.
Threatveil Oss
Open-source assurance infrastructure for tracking whether security evidence still applies as autonomous AI systems change.
Deterministic Agent Gatekeeper
Determinism you can prove for AI agent tool calls: a pure decision core whose every verdict re-derives hash-exact from its record.
Semantic Gps Hackathon
Mission control for AI agents. Governance gateway between AI agents and MCP-connected business systems: live policies, saga rollback, audit on every call, Tool Relationship workfl…
Purse
Purse is the enforcement point your agent's payments pass through. Route every spend through it. Set limits, require human approval over a threshold, and get a tamper-evident log…
Agent Skill Guard
Security auditor for AI agent skills — catches hidden-Unicode prompt injection, reviewer-subversion, data exfiltration & post-install drift that markdown-only vetters miss.
Verifiable Memory
Verifiable memory for AI agents — 0% hallucination, cited answers or honest abstention, provable forgetting, deterministic. MCP server. CPU-only, no GPU.
Praxis Aegis
Trust-layer companion for Praxis: policy guardrails and signing-aware controls.
LLM Redteam
面向中文场景的开源 LLM 红队测试工具 | An open-source red team toolkit for LLM applications, covering prompt injection, jailbreak and info leakage attacks, with OpenAI/Dify/MCP adapters and LLM-a…
Multiclaw Harness
Regression harnesses for agentic AI workflows: tool-call assertions, trajectory diffs, budget checks, security fixtures, and CI gates.
AI Workflow Integration Hub
FastAPI demo for routing customer events through AI-style workflow triage, integration issue detection, and Markdown report generation.