🎩 You're Invited:Meet the Socket team at Black Hat in Las Vegas, August 3-6.RSVP
Sign In

@agentguard47/mcp-server

Package Overview
Dependencies
Maintainers
1
Versions
4
Alerts
File Explorer

Advanced tools

Socket logo

Install Socket

Detect and block malicious and high-risk dependencies

Install

@agentguard47/mcp-server - npm Package Compare versions

Comparing version
0.2.0
to
0.2.1
+1
-1
dist/index.js

@@ -9,3 +9,3 @@ #!/usr/bin/env node

name: "agentguard",
version: "0.1.0",
version: "0.2.1",
});

@@ -12,0 +12,0 @@ let client;

{
"name": "@agentguard47/mcp-server",
"version": "0.2.0",
"description": "MCP server for AgentGuard — query traces, alerts, usage, and costs from AI coding agents",
"version": "0.2.1",
"description": "MCP server for coding-agent traces, alerts, costs, usage, and budget health",
"mcpName": "io.github.bmdhodl/agentguard47",
"license": "MIT",
"type": "module",
"repository": {
"type": "git",
"url": "https://github.com/bmdhodl/agent47.git",
"directory": "mcp-server"
},
"homepage": "https://github.com/bmdhodl/agent47/tree/main/mcp-server",
"bugs": {
"url": "https://github.com/bmdhodl/agent47/issues"
},
"keywords": [
"mcp",
"model-context-protocol",
"agentguard",
"coding-agents",
"agent-safety",
"runtime-guardrails"
],
"bin": {

@@ -8,0 +26,0 @@ "agentguard-mcp": "./dist/index.js"

+42
-13
# AgentGuard MCP Server
MCP (Model Context Protocol) server that connects AI coding agents to the AgentGuard Read API. Lets agents query their own traces, alerts, costs, and budget status.
MCP (Model Context Protocol) server that connects coding agents to the
AgentGuard Read API. It lets agents inspect their own traces, alerts, usage,
costs, and saved spend after the local SDK is already in place.
The boundary is deliberate:
- The SDK proves runtime enforcement locally
- The MCP server gives coding agents a narrow read surface over retained data
- The hosted dashboard stays the operational control plane
## Published Package
```bash
npx -y @agentguard47/mcp-server
```
## Tools

@@ -10,4 +23,4 @@

| `query_traces` | Search recent traces, filter by service/time range |
| `get_trace` | Get full event tree for a specific trace ID |
| `get_alerts` | Get guard alerts (loops, budget exceeded, errors) |
| `get_trace` | Get the full event tree for a specific trace ID |
| `get_alerts` | Get guard alerts such as loops, budget exceeded, and errors |
| `get_usage` | Check event quota usage and plan limits |

@@ -32,4 +45,4 @@ | `get_costs` | Get cost breakdown by model for the current month |

"agentguard": {
"command": "node",
"args": ["/path/to/agent47/mcp-server/dist/index.js"],
"command": "npx",
"args": ["-y", "@agentguard47/mcp-server"],
"env": {

@@ -43,15 +56,31 @@ "AGENTGUARD_API_KEY": "ag_your_key_here"

## Build & Run
Any other MCP-compatible client that can launch an npm package over stdio can
use the same command.
## Local Build & Run
```bash
npm ci # Install dependencies
npm run build # Compile TypeScript
npm start # Run server (stdio transport)
npm run dev # Watch mode for development
npm ci
npm run build
npm start
```
## Registry Readiness
This repo now includes official MCP registry metadata in
[`server.json`](server.json). The npm package is already public, so the
remaining registry work is metadata publication:
```bash
mcp-publisher login github
mcp-publisher publish
```
After that, verify the server is searchable from the MCP Registry API and then
submit it to downstream directories like Glama and `awesome-mcp-servers`.
## Architecture
- `src/index.ts` — Server entry point, tool registration (stdio transport)
- `src/client.ts` — HTTP client wrapping `/api/v1/` endpoints
- `src/tools.ts` — 6 MCP tool definitions and handlers
- `src/index.ts` - Server entry point, tool registration (stdio transport)
- `src/client.ts` - HTTP client wrapping `/api/v1/` endpoints
- `src/tools.ts` - 6 MCP tool definitions and handlers