MCP Server
Security

Your AI agents run MCP servers that can read files, execute commands, access databases, and send messages. Most are unsigned, unaudited, and invisible to your security tools.

Works with Claude Code, Cursor, Windsurf, VS CodeOpen source (Apache-2.0)No account required
View all CLI commands

MCP Servers Extend What Your AI Agent Can Do

The Model Context Protocol (MCP) lets AI agents connect to external servers that provide tools beyond built-in capabilities. A developer installs an MCP server and their AI agent gains the ability to run shell commands, query databases, process payments, or send messages on their behalf.

MCP servers are configured in JSON files scattered across your machine — in Claude Code settings, Cursor configs, Windsurf directories, and VS Code extensions. Each configuration grants new capabilities to the agent. Most organizations have no inventory of which MCP servers are installed, what they can do, or whether they have been tampered with.

Security teams face a supply chain problem: MCP servers are installed from npm, PyPI, or GitHub with no signing, no attestation, and no centralized visibility. A compromised or malicious MCP server has the same access as the AI agent itself.

No signing
Most MCP servers have no cryptographic identity or integrity verification
No visibility
Configs are scattered across multiple tools with no centralized inventory
No audit
Capabilities are granted implicitly with no review or approval workflow

See What Your MCP Servers Can Actually Do

OpenA2A CLI analyzes each MCP server configuration and surfaces its capabilities in plain language.

Can run commands on your computer

shell-access

Can read and modify databases

database

Can access payment systems

payments

Can send messages on your behalf

messaging

Can read files across your filesystem

filesystem

Can make network requests to external APIs

network

Three Commands. Full MCP Security.

Discover, audit, and sign your MCP server configurations from the terminal.

Discover

Find every MCP server on your machine

Terminal
$ opena2a detect

MCP Server Discovery
  Scanning: project-local + machine-wide configurations

  Claude Code    3 servers found
  Cursor         1 server found
  VS Code        2 servers found

  Total: 6 MCP servers across 3 platforms

  Server            Platform      Transport   Status
  filesystem        Claude Code   stdio       Running
  postgres-mcp      Claude Code   stdio       Running
  stripe            Cursor        sse         Running
  slack-bot         VS Code       stdio       Stopped
  github-mcp        VS Code       stdio       Running
  custom-tools      Claude Code   stdio       Running

Audit

Analyze capabilities and trust scores

Terminal
$ opena2a mcp audit --registry

MCP Security Audit
  Server: stripe
  Platform: Cursor
  Transport: sse

  Capabilities detected:
    [payments]   Can access payment systems
    [network]    Can make network requests to external APIs

  Registry Trust Score: 50/100
  Signed: No
  Last verified: Never

  Recommendations:
    - Sign this server with: opena2a mcp sign stripe
    - Review payment scope permissions
    - Enable transport encryption for SSE connections

Sign

Cryptographic identity for tamper detection

Terminal
$ opena2a mcp sign stripe

MCP Server Signing
  Server: stripe
  Algorithm: Ed25519

  Configuration hash:  a3f8c2d1...
  Signature:           MCowBQYDK2Vw...
  Attestation saved:   ~/.opena2a/attestations/stripe.json

  Next verification will detect any configuration changes.
  Re-run 'opena2a mcp audit' to confirm signed status.

Community Trust Scores from the Registry

The --registry flag enriches your local scan results with community trust data from the OpenA2A registry. See how the broader community rates each MCP server.

filesystem
Trust: 72/100
Signed
postgres-mcp
Trust: 65/100
Signed
stripe
Trust: 50/100
Unsigned
github-mcp
Trust: 78/100
Signed
slack-bot
Trust: 45/100
Unsigned
custom-tools
Not in registry
Unsigned

Trust scores are computed from community reports, signing status, maintainer reputation, and vulnerability history.

Enterprise Asset Inventory and Reporting

Export MCP server inventories for compliance, asset management, and executive reporting.

CSV Export

Asset inventory for CMDB integration

opena2a detect --format csv > mcp-inventory.csv

Import into ServiceNow, Jira Assets, or any CMDB. Includes server name, platform, transport, capabilities, and trust score.

HTML Report

Executive summary for stakeholders

opena2a mcp audit --format html > report.html

Visual report with trust score distribution, unsigned server count, capability breakdown, and remediation priorities.

Start Auditing Your MCP Servers

One command. No account required. See every MCP server across Claude Code, Cursor, Windsurf, and VS Code.

npx opena2a-cli detect --registry

Get the MCP Security Guide

A practical guide to auditing, signing, and governing MCP servers across your organization. Delivered to your inbox.

Stay Updated on AI Agent Security

Weekly insights, vulnerability alerts, and best practices