Search: ai-safety
Search MCP servers and agent skills by name, description, category or topic — 25 results.
jnMetaCode/shellward
AI Agent Security Middleware & MCP Server with 8-layer defense including prompt injection detection, DLP data flow tracking, command blocking, and PII detection. 7 MCP tools, zero dependencies.
IgorGanapolsky/ThumbGate
MCP server that blocks AI coding agents from repeating mistakes — turns thumbs-up/down feedback into enforced pre-action gates via PreToolUse hooks. Install: `npx thumbgate`.
IgorGanapolsky/mcp-memory-gateway
Pre-action gates that prevent AI coding agents from repeating known mistakes. Captures explicit feedback, auto-promotes failures into prevention rules, and enforces them via hooks.
sekera-radim/impri
Human-in-the-loop approval inbox for AI agents. An agent submits a proposed action (send email, post comment, run a command) via `impri_push_action`, a human approves, rejects, or edits it from a web, mobile, or Slack/Discord/Telegram inbox, and the agent only proceeds on an approved decision. The gate is a data dependency, not a prompt. Full audit trail, self-hostable (MIT, Docker Compose). `npx
agentkitai/agentlens
Tamper-evident observability for AI agents: a SHA-256 hash-chained audit log with chain verification and signed export (EU AI Act Art. 12). Instrument any agent with zero code via `npx -y @agentlensai/mcp`; also ingests OpenTelemetry GenAI traces.
sint-ai/sint-protocol
Security-first MCP governance proxy (`sint-mcp`) with capability tokens, T0-T3 approval tiers, fail-closed execution, and tamper-evident audit receipts. Includes a separate `sint-scan` CLI for preflight MCP tool-risk audits.
Govcraft/rust-docs-mcp-server
Provides up-to-date documentation context for a specific Rust crate to LLMs via an MCP tool, using semantic search (embeddings) and LLM summarization.
AperionAI/shield
Local guardrail proxy for AI coding agents. Wraps any MCP server (stdio or Streamable HTTP) and blocks destructive tool calls — DROP TABLE, rm -rf, force-push — before they execute. MCP supply-chain protection: TOFU tool-catalog pinning against rug pulls, plus tool-description and tool-result scanning for tool poisoning and prompt injection. 51 starter rules, approval gates, audit logging. Single
kiro0x/five-mcp
LLM character consistency engine — generates structured JSON constraints from 4 multiple-choice questions about an AI's psychology. Drop the JSON into any LLM's system prompt to prevent persona drift; reduces inference cost from retries. 160,000 personality patterns; works with any LLM.
lacs-project/sysknife
Security-hardened MCP server for Linux system administration via 189 typed actions instead of shell strings, with an Ed25519-signed hash-chain audit log, one-time TTL approval receipts, and automatic rollback. Works with Claude Code, Cursor, and Codex CLI.
babyblueviper1/invinoveritas
Lightning-native AI reasoning, decisions, persistent memory, and agent marketplace for autonomous agents. Pay-per-use via Bitcoin Lightning. Register free — 250 starter sats. Agents earn sats selling services (seller keeps 95%), DM each other, and run autonomously. `npm install invinoveritas-mcp`
bmdhodl/agent47
Runtime guardrails and incident read access for coding agents. Query AgentGuard traces, alerts, usage, costs, and budget health.
behrensd/mcp-firewall
Deterministic security proxy (iptables for MCP) that intercepts tool calls, enforces YAML policies, scans for secret leakage, and logs everything. No AI, no cloud.
knowledgepa3/gia-mcp-server
Enterprise AI governance layer with 29 tools: MAI decision classification (Mandatory/Advisory/Informational), hash-chained forensic audit trails, human-in-the-loop gates, compliance mapping (NIST AI RMF, EU AI Act, ISO 42001), governed memory packs, and site reliability tools.
Acacian/aegis
Policy-based governance for AI agent tool calls. YAML policies, approval gates, risk assessment, and audit logging. Cross-platform: LangChain, OpenAI, Anthropic, MCP.
Chimera-Protocol/csl-core
Deterministic AI safety policy engine with Z3 formal verification. Write, verify, and enforce machine-verifiable constraints for AI agents via MCP.
loicfontaine-max/qorami-sdk
Check an email before an AI agent sends it: returns send / ask-a-human / block, with machine reason codes and prompt-injection detection.
airblackbox/air-blackbox-mcp
EU AI Act compliance scanner for Python AI agents. Scans, analyzes, and remediates LangChain/CrewAI/AutoGen/OpenAI code across 6 articles with 10 tools including prompt injection detection, risk classification, and trust layer integration. The only MCP compliance server that generates fix code, not just findings.
ark-forge/mcp-eu-ai-act
EU AI Act compliance scanner that detects regulatory violations in AI codebases with risk classification and remediation guidance.
wei9072/aegis
AI-agent admission-control MCP server: validates file edits against Ring 0 syntax + Ring 0.5 structural-cost regression + workspace boundary (path / glob / shell-redirect / symlink). Negative-space framing — emits BLOCK / WARN / PASS verdicts, never coaches the agent.
cuttalo/depscope
Package Intelligence for AI agents. 22 tools across 17 ecosystems (npm/pypi/cargo/go/maven/nuget/rubygems/composer/pub/hex/swift/cocoapods/cpan/hackage/cran/conda/homebrew) — check health, vulnerabilities (OSV + CISA KEV + EPSS), typosquats, malicious flags, alternatives, known bugs, breaking changes, stack compatibility and error-to-fix. 31k+ packages, 2.2k+ CVEs enriched. Zero auth, MIT. Remote
sim-xia/blind-auditor
A zero-cost MCP server that forces AI to self-correct generation messages using prompt injection, independent self-audition and context isolation.
bluetieroperations-create/blackwall-mcp
Pre-action risk gate for AI agents. One `forecast` tool the agent calls before any irreversible action (send money, run SQL, delete data); returns a risk score (0–100), reversibility class, named red flags from 28 failure modes, and a gate: proceed / confirm / human-required.
imran-siddique/agentos-mcp-server
Agent OS MCP server for AI agent governance with policy enforcement, code safety verification, multi-model hallucination detection, and immutable audit trails.