traceguard-mcp
TraceGuard MCP Server —— 让任何支持 MCP 的客户端(Claude Desktop、Cursor、OpenCode、Claude Code) 在执行工具之前先问一句「这个动作允许吗」。
零依赖,本地运行,不联网。
安装
pip install traceguard-mcp
# 想用你自己的策略文件,再装核心:
pip install ai-traceguard
它暴露什么
| 工具 | 作用 |
|---|---|
traceguard_check(tool_name, args) |
判断该工具调用是否被允许,返回 allowed / risk / reason |
traceguard_report() |
审计摘要(调用数 / 拦截数 / 链是否完好) |
配置到 MCP 客户端
{
"mcpServers": {
"traceguard": {
"command": "traceguard-mcp"
}
}
}
或用 Python 模块方式:
{
"mcpServers": {
"traceguard": {
"command": "python",
"args": ["-m", "traceguard_mcp.server"]
}
}
}
环境变量
| 变量 | 说明 |
|---|---|
TRACEGUARD_MCP_POLICY |
策略文件路径(默认 policies/base.yaml,需装 ai-traceguard) |
TRACEGUARD_MCP_DB |
审计数据库路径(默认 traceguard-mcp.db) |
怎么用
在 Agent 的工具分发循环里,执行任何有副作用的工具之前先调 traceguard_check:
res = mcp_call("traceguard_check", {"tool_name": name, "args": args})
if not res["allowed"]:
return f"⛔ 被拦下:{res['reason']}" # 把原因回传给模型
# 放行,执行真实工具
把
reason回传给模型,它会自己改用安全做法,体验远好于直接报错。
实测
initialize -> traceguard v0.1.0(协议 2024-11-05)
tools/list -> traceguard_check, traceguard_report
check delete_file -> {"allowed": false, "risk": "HIGH", "reason": "删除文件操作,策略阻断"}
check read_file -> {"allowed": true, "risk": "LOW"}
许可证
MIT
Metadata
Release files for traceguard-mcp 0.1.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| traceguard_mcp-0.1.0.tar.gz | 4.5 kB | Without provenance | Details for traceguard_mcp-0.1.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| traceguard_mcp-0.1.0-py3-none-any.whl | Python 3 | none | any | Details for traceguard_mcp-0.1.0-py3-none-any.whl |
Total release size: 10.1 kB
Release files / traceguard_mcp-0.1.0.tar.gz
| Download URL | traceguard_mcp-0.1.0.tar.gz |
|---|---|
| Size | 4.5 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
026d5a818ee70074a794f1b61aa32e83d52eca04b5d57414ebc540c017ab9531
|
|
BLAKE2b-256 checksum How to use checksums |
6669d69edfe46ad5167b25fa77391f9070e6f973b5a3f3668cdb16186d85ead2
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.13.2
|
Release files / traceguard_mcp-0.1.0-py3-none-any.whl
| Download URL | traceguard_mcp-0.1.0-py3-none-any.whl |
|---|---|
| Size | 5.6 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
141c9568d48bd285645571de956a2ffbedf6d3797889a1094425922f8f6821d8
|
|
BLAKE2b-256 checksum How to use checksums |
ffb14e660859fdfec072eb3e2b1dac09e8334290cbc9228138639e0e4b495250
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.13.2
|
Security for traceguard-mcp 0.1.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.