Agentaegis MCP

What astafford8488/agentaegis-mcp MCP server does

The astafford8488/agentaegis-mcp MCP server exposes cybersecurity operations as structured MCP tools for AI agents. Its coverage spans five main areas: compliance, vulnerability management, code security, defensive operations, and identity. It can assess SOC 2, ISO 27001, HIPAA, PCI-DSS, and NIST CSF controls; create evidence plans and audit reports; scan networks and web applications; look up CVEs; audit TLS, DNS, and email security; and review access or MFA coverage.

The trust-oriented tools address risks specific to agent ecosystems. scan_mcp_plugin checks an MCP server or plugin before installation or trust, scan_skill examines an agent skill for prompt-injection and excessive permissions, and vet_endpoint returns a PROCEED, CAUTION, or BLOCK verdict before an agent calls or pays an endpoint. Code-focused tools include static analysis, secret detection, and dependency auditing. Threat intelligence and incident triage tools support defensive investigations.

How it works

The project supports two transport modes. Local stdio mode runs the server process directly for a client such as Claude Desktop. Remote deployments use Streamable HTTP at /mcp; the repository documents local HTTP development and production deployment with Docker Compose or Railway.

Internally, the server presents MCP tool definitions and delegates work to open-source engines including nmap, Nuclei, sslyze, Semgrep, trufflehog, and trivy. Other tools query services such as NVD, AbuseIPDB, AlienVault OTX, abuse.ch, Have I Been Pwned, and Shodan. The architecture also includes target validation, isolated scan processes, rate limits, logging, and Supabase-backed customer, key, job, usage, and webhook data.

Setup and configuration

To run astafford8488/agentaegis-mcp locally, clone the repository, install dependencies with pnpm, copy .env.example to .env, build the project, and start the stdio entry point. The documented Claude Desktop configuration launches dist/index.js with Node and passes security-service API keys through the client environment.

Remote operation requires deployment configuration. The documented production variables include Supabase connection details, an x402 payee wallet, and API keys for NVD, AbuseIPDB, AlienVault OTX, abuse.ch, and HIBP. The HTTP API also exposes unauthenticated health and pricing routes, customer creation and API-key issuance routes, authenticated usage and job-status routes, and the MCP endpoint.

Remote authentication supports a database-backed API key or x402 micropayments. API-key customers use a monthly limit or prepaid billing model. x402 clients receive payment requirements and retry with an X-PAYMENT header after signing a USDC payment on Base.

Tools and capabilities

The catalog includes 28 tools, with paid operational tools and free discovery or account tools. Representative capabilities include:

• Evaluate compliance frameworks and generate policies, evidence plans, gap analyses, and audit reports.
• Scan network targets and web applications, then prioritize vulnerabilities or retrieve CVE data.
• Inspect source code, secrets, and dependencies without executing cloned repositories.
• Investigate indicators, triage incidents, and audit DNS, email, access, and MFA controls.
• Assess MCP plugins, agent skills, and external endpoints before installation or use.

Limitations and notes

Use scanning tools only against systems you own or are explicitly authorized to test. Target validation permits public IP addresses and registered domains while blocking private or reserved ranges. Scans are limited by concurrency and per-target rate limits, and code repositories are shallow-cloned with size and timeout limits.

Secret scanning does not return complete secret values; it redacts them to the first and last four characters. Tool calls are logged with 90-day retention. The README states that live pricing is available from the deployed /pricing endpoint, so costs can change from the examples in the repository.