Safety Fence MCP
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Safety Fence MCPGuard this outbound payload and redact sensitive data"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Safety Fence MCP
这是验收版 SKILL.md 的本地 MCP Server 最小实现。它通过唯一工具
guard_payload 对本地任务或即将外发的精确 payload 做 ALLOW、
REWRITE、REFUSE 判定。
当前策略基线:
sha256:24fdc1b886003e5a76cfc8e8103bbcc5afead5c0f9d3af0167462bf76055e3d7安全边界
Server 使用本地
stdio,无网络依赖、无外部模型调用、无遥测。ALLOW不复制原始 payload;REWRITE只返回安全 payload;REFUSE不返回原文。日志接口只包含决策、原因代码、目标类型、字节数和脱敏数量。
MCP 本身是决策面,不是透明代理。OpenClaw 必须在可信外发边界强制调用并执行结果。
围栏超时、进程异常或返回无效结构时,调用方必须 fail-closed,禁止本次外发。
Related MCP server: mcp-doorman
启动
无需安装第三方依赖:
python -m safety_fence也可以安装为本地包后使用命令:
python -m pip install -e .
safety-fence-mcpMCP 客户端配置见 mcp-config.example.json。cwd 必须替换为本项目绝对路径。
工具输入
{
"flow": "EGRESS",
"task_context": "用云端模型总结随访重点",
"payload": {
"message": "患者韩珂,手机号 13500006666"
},
"destination": {
"kind": "cloud_model",
"name": "remote-model"
}
}flow 必须由可信拦截器填写,不能采用模型自行声明的值。EGRESS 必须提供
destination.kind。payload 支持任意 JSON,默认最大 2 MiB、最大嵌套 32 层。
返回值同时提供 MCP 文本兼容内容和 structuredContent:
{
"policy_version": "sha256:...",
"decision": "REWRITE",
"reason_codes": ["PHONE", "ROLE_BOUND_NAME"],
"reason": "敏感信息属于附带上下文,已最小化处理。",
"rewritten_payload": {
"message": "患者<REDACTED>,手机号 <REDACTED>"
},
"user_notice": "为保护隐私与安全,已使用脱敏或最小化后的 payload 继续。",
"redaction_count": 2
}强制执行
safety_fence.interceptor.EgressInterceptor 是最小参考适配器:
ALLOW:把原始 payload 交给 sender。REWRITE:只把rewritten_payload交给 sender。REFUSE:不调用 sender。引擎或审计异常:不调用 sender,并抛出
FenceUnavailable。
生产接入时,应把同样的状态机放到云模型、远程工具、HTTP API、邮件、聊天、 仓库、远程日志和 Telemetry 的共同出口,而不是让 Agent 自主决定是否调用围栏。
验证
python -m unittest discover -s tests -v测试覆盖验收版锚点、嵌套 JSON 脱敏、改写幂等、MCP 生命周期、真实 stdio 子进程往返、原文不外发以及失败关闭。
当前最小版的范围
当前语义层是本地确定性意图/角色分类器,覆盖 SKILL.md 中的验收锚点和核心
不变量,不依赖云端 LLM。后续如引入本地模型,应作为保守规则之后的语义补充,
且仍必须通过相同 Golden Tests、脱敏复检和失败关闭机制。
This server cannot be deployed
Maintenance
Related MCP Connectors
Guarded MCP server for agent-readable business truth, provenance, readiness, and discovery.
An MCP server that provides Javelin Standalone Guardrails
- gatewayOAuthai.sealgate
MCP gateway with runtime security policy, tool-call-level control, and audit of agent actions.
MCP server for detecting and redacting PII (Personally Identifiable Information) in PDF documents.
Related MCP Servers
- AlicenseNot gradedqualityBmaintenanceMCP server for AI agent security guardrails. Provides input validation, prompt injection detection, PII redaction, output filtering, policy enforcement, rate limiting, and comprehensive audit logging.31 npm1MIT
- AlicenseNot gradedqualityCmaintenanceA drop-in proxy that guards MCP servers with policy enforcement, secret redaction, prompt-injection screening, rug-pull detection, rate limiting, and audit logging.5 npmApache 2.0

context-firewallofficial
AlicenseNot gradedqualityCmaintenanceA local MCP server that acts as a policy-based firewall for personal context, letting users control agent access to their data with deny-by-default rules, per-field grants, and an audit trail.38 npmApache 2.0- AlicenseNot gradedqualityCmaintenanceA fail-closed policy boundary that translates local stdio MCP clients to authenticated Streamable HTTP servers, enforcing allowlists or read-only modes and redacting credentials from audit trails.Apache 2.0