Your MCP Tools
Are a Backdoor
The Model Context Protocol is the standard way AI coding tools talk to external services. When you use Claude Code, Cursor, or Windsurf with a filesystem server, a database connector, or any of the 8,600+ MCP servers listed on public directories. Every action goes through MCP.
The AI sends a JSON-RPC request like tools/call with a tool name and arguments. The MCP server executes it. Read a file, run a shell command, query a database. Whatever the agent asks.
You have a filesystem MCP server configured. Claude Code is helping you refactor a project. Normal workflow. The AI reads your source files, checks your package.json, looks at your test suite.
Your SSH private key.
The server executed it like any other read. No distinction between a project file and your most sensitive credential.
Existing protections operate at the wrong layer.
Binary allow/deny by tool. If you allow read_file, you allow all reads. You can’t say “allow project files but block .ssh/.” No argument inspection.
Checks tool descriptions at install time. In one academic study, detected 4 of 120 poisoned servers, a 3.3% detection rate. Scanners are a useful first layer, but runtime enforcement is needed too.
Some tools route your tool calls through external APIs. Your code and secrets leave your machine. For privacy-sensitive work, local-only enforcement is the safer default.
Same scenario. Same MCP server. But now mcpwall sits between the AI tool and the server, intercepting every JSON-RPC message.
The rule that caught the SSH key theft:
Eight default rules cover the most common attack vectors out of the box: SSH keys, .env files, credential stores, browser data, destructive commands, pipe-to-shell, reverse shells, and secret leakage.
No config needed. The defaults apply automatically.
Then change your MCP config:
Or let mcpwall find and wrap your servers automatically:
Critical RCE in mcp-remote. 437K+ installs affected.
Major enforcement provisions take effect.
MCP adoption is accelerating. It’s been donated to the Linux Foundation, every major AI coding tool supports it, and the server ecosystem is growing by hundreds per week. The attack surface is growing faster than the security tooling. If you use MCP servers, a programmable policy layer between your AI agent and those servers is defense in depth.