Tutorial · Available

Protect your MCP tools

Put the proxy between your agent and its MCP servers, then point the agent at it.

Status: Available. This integration is built and tested. Pilot companies receive the Guardrino package from HisnLabs.

Developer reference · Install guide · AI platforms · Open model card · OpenAI-compatible proxy · Streaming · LiteLLM · MCP security proxy · Tutorial: OpenAI-compatible app · Tutorial: LiteLLM

Try it today Guardrino-t7 Open, an earlier version of our classifier, is public on Hugging Face under Apache-2.0. The demo runs in your browser: what you type stays on your device. Live demo · Model card · Same demo on Hugging Face

Step 1 · List your MCP servers

Tell the proxy which MCP servers to sit in front of, in the policy file. Each one gets a name, and the agent reaches it at /mcp/<name>.

[mcp]
on_tool_poisoning = "drop"   # "drop": remove a flagged tool from the list | "refuse": refuse the whole list
timeout_secs = 120
servers = [
  { name = "files", url = "https://mcp.example/mcp", headers = { Authorization = "Bearer UPSTREAM_TOKEN" } },
]

Step 2 · Point your agent at the proxy

In the agent’s MCP settings, replace each server’s address with the proxy’s address.

{
  "mcpServers": {
    "files": { "url": "https://guardrino.example/mcp/files" }
  }
}

Local stdio server? Run guardrino-mcp-stdio --server-cmd "<command>" in its place instead. No servers entry is needed.

{
  "mcpServers": {
    "files": { "command": "guardrino-mcp-stdio", "args": ["--server-cmd", "npx some-mcp-server"] }
  }
}

Step 3 · See what is checked

When tools are listed, their descriptions are checked.
When a call goes out, its arguments are checked for the canary token, personal data and URLs.
When a result comes back, it is read as context, in windows.

To use the canary, create one with POST /v1/canary and put it in the agent’s system prompt.

Request early access