Guardian MCP runs on your own machine — the client starts it, so there is no endpoint to ping. Last commit 1 Nov 2026.
Routes agent tool calls through Guardian for ALLOW/ASK/DENY before execution.
Today is the operative word: we check Guardian MCP every 15 minutes and re-read its code on every release. Watch it and you find out the day that stops being true.
This server runs on your own machine — install it with the package manager and the client starts it for you. Package name taken from the official registry entry.
claude mcp add guardian-mcp -e 'GUARDIAN_BASE_URL=<GUARDIAN_BASE_URL>' -e 'GUARDIAN_AGENT_CREDENTIAL=<GUARDIAN_AGENT_CREDENTIAL>' -e 'GUARDIAN_CONNECTION_ID=<GUARDIAN_CONNECTION_ID>' -- npx -y @odingard/guardian-mcp
{
"mcpServers": {
"guardian-mcp": {
"command": "npx",
"args": [
"-y",
"@odingard/guardian-mcp"
],
"env": {
"GUARDIAN_AGENT_CREDENTIAL": "<GUARDIAN_AGENT_CREDENTIAL>",
"GUARDIAN_BASE_URL": "<GUARDIAN_BASE_URL>",
"GUARDIAN_CONNECTION_ID": "<GUARDIAN_CONNECTION_ID>"
}
}
}
}
[mcp_servers.guardian-mcp]
command = "npx"
args = ["-y", "@odingard/guardian-mcp"]
env = { GUARDIAN_BASE_URL = "<GUARDIAN_BASE_URL>", GUARDIAN_AGENT_CREDENTIAL = "<GUARDIAN_AGENT_CREDENTIAL>", GUARDIAN_CONNECTION_ID = "<GUARDIAN_CONNECTION_ID>" }
{
"mcpServers": {
"guardian-mcp": {
"command": "npx",
"args": [
"-y",
"@odingard/guardian-mcp"
],
"env": {
"GUARDIAN_AGENT_CREDENTIAL": "<GUARDIAN_AGENT_CREDENTIAL>",
"GUARDIAN_BASE_URL": "<GUARDIAN_BASE_URL>",
"GUARDIAN_CONNECTION_ID": "<GUARDIAN_CONNECTION_ID>"
}
}
}
}
{
"servers": {
"guardian-mcp": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@odingard/guardian-mcp"
],
"env": {
"GUARDIAN_AGENT_CREDENTIAL": "<GUARDIAN_AGENT_CREDENTIAL>",
"GUARDIAN_BASE_URL": "<GUARDIAN_BASE_URL>",
"GUARDIAN_CONNECTION_ID": "<GUARDIAN_CONNECTION_ID>"
}
}
}
}
Put your own values in place of
<GUARDIAN_BASE_URL> (Guardian control-plane base URL), <GUARDIAN_AGENT_CREDENTIAL> (Agent credential (gac1.…)), <GUARDIAN_CONNECTION_ID> (Default tool connection id for invocations).
The author’s registry entry marks them as required.
Optional settings from the registry entry, left out of the block above:
GUARDIAN_ADAPTER_ID (Optional adapter override).
Pre-execution governance for AI agents. Validates tool calls before they execute.
Safety execution proxy protecting clients during tool calls.
Explainable ALLOW/WARN/BLOCK risk decisions for AI agent actions before execution, signed as OAA.
Security gateway for MCP agents: blocks prompt-injection-driven tool calls before they execute.
Pre-execution gate for AI-agent tool calls. Touch ID approval. Tamper-evident audit log.
MCP-native router and multi-agent control plane: plan, route tool calls, stream execution.
Find the best eligible service for an agent task through one task-first MCP tool.
Scan prompts, tool definitions and model output for injection, and guard agent tool calls.
Answers built from our own checks of this server.