Security MCP Servers

One local gate for Claude Code, Codex, Copilot CLI and Antigravity: allow, deny and ask rules over MCP and built-in tool calls, approvals from a terminal UI, and hash-chained receipts.

Claim-level hallucination detection with sources, agent trajectory verification, risk gating for computer-use actions and prompt-injection defense.

Fail-closed approval layer for unattended agent runs: checks actions against org policy, requests human approval and logs every decision to a causal ledger.

Install a large security toolset and discover and run it through an authorization-gated server for CTF, pentesting, bug bounty and DFIR.

Turn other AIops tools' audit trails into tamper-evident, hash-chained evidence bundles mapped to HIPAA, PCI-DSS, SOC 2 and GDPR. Offline and deterministic.

Self-hosted OWASP ZAP integration with guided security scans, findings summaries and report generation.

Payment guardrails for AI agents: spend caps, budget and approval gates, prompt-injection defense, an encrypted secret vault and an audit trail.

Self-hosted security layer that evaluates AI agents' proposed blockchain transactions and returns an explainable allow, warn or block decision.

EVM bytecode capability analysis, EIP-1967 proxy resolution, gas recommendations and preflight safety checks for agents on Base.

Malcolm threat hunting: search and aggregate network traffic, query Suricata alerts, browse Arkime sessions and resolve NetBox assets. Read-only by default.

CyberChef MCP server providing 28 core data transformations, ciphers, hashing, JWT decoding, representation-calibrated Shannon entropy, and multi-layer payload deobfuscation for AI coding and security agents. npx -y @noorfatima123456/cyber-chef-mcp

Read-only safety checks for Base and EVM: honeypot and rug-pull detection, risky approvals, verified source, balances, ENS and Basenames, gas and prices.

Transparent MCP proxy that keeps a signed, hash-chained audit trail of agent tool calls, with YAML policy enforcement and OpenTelemetry export.

JSON risk briefs for wallet addresses, tokens and URLs: scores, flags and sources from phishing, scam and URL-reputation heuristics, including OpenPhish.

Local, zero-trust masking of PII and secrets.

Supply-chain gate for agent skills and MCP servers: detects poisoned tool definitions, pins servers by hash, checks drift in CI and proxies only pinned tools.

Check lockfiles against malicious-package advisories and scan artifacts for credential theft, exfiltration, obfuscation, prompt injection and droppers.

DNS and email security scans via IntoDNS.ai: SPF, DKIM, DMARC, DNSSEC, MTA-STS, BIMI, TLS, blacklist and deliverability checks plus security header analysis.

Fail-closed trust gate for agents: fact-checking, prompt-injection detection, content moderation, PII and secret detection, and a pre-action guardrail.

Scan MCP servers and skills before install, vet endpoints before calling them, and run vulnerability scans, threat intel and compliance checks.

Deterministic audit layer that checks LLM and agent outputs against jailbreak, safety and trading compliance policies and logs verdicts to a hash chain.

Register and verify domains, queue security scans (headers, TLS, DNS, ports, CVEs and AI-specific checks) and read findings.

Verify IBAN, VAT, VIN, barcodes, national and tax IDs, crypto addresses, phone numbers, dates and holidays against checksums and curated data.

Domain security checks: DNS, DNSSEC, TLS grading, HTTP security headers, SPF, DKIM and DMARC, port scans, ASN, RDAP and WHOIS, plus an overall health score.

Agent compliance and identity: verify an agent's X.509 certificate scope, read its behavioral risk score and append to an immutable audit trail.

Honeypot threat intelligence from a sensor network: IP reputation, scanner classification, CVE probing trends and JA4, JA4H and HASSH fingerprints.

Infrastructure diagnostics: DNS checks, TLS certificate chain inspection, email security, global DNS propagation and DNS latency benchmarking.

Cross-repository code knowledge graph for Java, Kotlin, JavaScript and TypeScript: call-flow tracing, taint analysis, reachability and license compliance.

Least-privilege compiler for agents: record real tool calls, compile a minimal policy, enforce it at an MCP gateway and keep a hash-chained audit.

SSRF-hardened web gateway with web fetch, web search and Metis verification, sharing an audited security core with Metis and ARGUS.

Verifiable oracle tools via AIMarket Hub: Platon VRF randomness, Chronos VDF computation and verification, and LUMEN reputation scores.

Query ARGUS-3 and check its status; WARDEN vets third-party MCP servers with a scored firewall, tool-definition pinning and drift detection.

Cryptographic provenance, blind reputation scoring and tamper-evident logging for agent interactions, with dispute resolution and trust-based matchmaking.

Redacts emails, phone numbers, national IDs, private keys and financial data before text reaches LLMs, in English, Spanish, Portuguese, German and Japanese.

URL security signals: redirects, brand impersonation, domain age, SSL validation, parked-domain detection, structural analysis and DNS enrichment.

Ghidra plugin for binary analysis: function inspection, decompilation, memory exploration and import/export analysis.

Real-time security framework for MCP servers that detects and blocks malicious agent behavior by analyzing tool call patterns and intent.

Zero-knowledge password manager and MCP credential gateway: policy-gated agent access, deterministic passwords, BIP-39 recovery and a hash-chain audit trail.

Policy-based governance for agent tool calls across LangChain, OpenAI, Anthropic and MCP: YAML policies, approval gates, risk assessment and audit logging.

Local-first MCP gateway for coding agents that redacts detected secrets from file reads and command output before they reach model context.

Reverse engineering tools from Solvitor: extract IDL files from closed-source Solana smart contracts and decompile them.

Agent identity, scoped delegation and governance: issue passports, verify narrowing-only delegation chains, enforce gateway policy and emit signed records.

Verify a photo's capture time and provenance: C2PA Content Credentials validation, EXIF/XMP consistency checks and pixel forensics combined into one verdict.

Query AgentForge Trust Scores for MCP servers before connecting, covering security, code health, behavioral audit, community trust and EU compliance.

Signed, recomputable safety scores for MCP servers, packages and tools, with offline-verifiable attestations and a GitHub Action to gate CI merges.

Identity and credential broker for MCP servers: issues scoped, short-lived credentials per agent, with audit logging and human approval gates.

Local guardrail proxy for MCP servers: blocks destructive tool calls, pins tool catalogs against rug pulls and scans for tool poisoning and prompt injection.

Proof-of-behavior enforcement for AI agents: define permit/forbid/require rules, enforce them at runtime and keep hash-chained, tamper-evident audit logs.