Any MCP Client or Server

Agent Wall works with every MCP server and every MCP client that uses stdio transport. No code changes needed.

Why it works universally

Agent Wall operates at the protocol level. The MCP protocol uses JSON-RPC 2.0 over stdio with newline-delimited messages. Agent Wall:

1. Spawns your MCP server as a child process
2. Intercepts the stdin/stdout byte stream
3. Parses JSON-RPC messages
4. Evaluates tools/call requests against your policy
5. Passes everything else through untouched

It doesn't know or care what the server does — it only sees the protocol messages. So whether you're running a filesystem server, a Stripe server, a custom Python server, or any of the 1000+ MCP servers available, Agent Wall protects it the same way.

Any MCP server

# Official MCP servers
agent-wall wrap -- npx @modelcontextprotocol/server-filesystem /path
agent-wall wrap -- npx @modelcontextprotocol/server-github
agent-wall wrap -- npx @modelcontextprotocol/server-playwright
agent-wall wrap -- npx @modelcontextprotocol/server-slack

# Popular third-party servers
agent-wall wrap -- npx @supabase/mcp-server-supabase
agent-wall wrap -- npx @stripe/mcp-server-stripe
agent-wall wrap -- npx mcp-server-postgres postgresql://localhost/db

# Servers written in any language
agent-wall wrap -- node my-server.js
agent-wall wrap -- python my_server.py
agent-wall wrap -- go run ./cmd/mcp-server
agent-wall wrap -- cargo run --bin my-mcp-server

# Docker-based servers
agent-wall wrap -- docker run -i my-mcp-server

Any MCP client

The integration pattern is always the same: replace the server command with agent-wall wrap -- <original-command>.

Supported clients (with auto-detection)

These clients are auto-detected by agent-wall scan and agent-wall doctor:

Client	Config Location
Claude Code	~/.claude/mcp_servers.json
Claude Desktop	Platform-specific (see Claude Desktop guide)
Cursor	~/.cursor/mcp.json
VS Code / Copilot	.vscode/mcp.json
Windsurf	~/.codeium/windsurf/mcp_config.json
Cline	~/.cline/mcp_settings.json
Continue.dev	~/.continue/config.json

Other MCP clients

For any MCP client not listed above, the integration is the same. Find where the client stores its MCP server config and change the command:

Before:

{
  "command": "npx",
  "args": ["some-mcp-server", "--flag"]
}

After:

{
  "command": "agent-wall",
  "args": ["wrap", "--", "npx", "some-mcp-server", "--flag"]
}

You can also scan a custom config file:

agent-wall scan --config /path/to/your/mcp-config.json

FAQ

Do I need to update Agent Wall when new MCP servers are released?

No. Agent Wall works at the protocol level. New MCP servers use the same JSON-RPC 2.0 protocol, so they're automatically supported.

The scan command has risk detection patterns for 48 known server types (for reporting purposes), but the wrap command and policy engine work with any server regardless.

Does it work with remote MCP servers (SSE/HTTP)?

Currently, Agent Wall supports stdio transport only (which is what most MCP clients use). Remote server support (SSE/HTTP) is planned for a future release.

Does it work with servers that use environment variables?

Yes. When you wrap a server, the child process inherits the parent's environment. You can also set env vars in your MCP client config:

{
  "command": "agent-wall",
  "args": ["wrap", "--", "npx", "mcp-server-postgres"],
  "env": {
    "DATABASE_URL": "postgresql://localhost/mydb"
  }
}