ThreatLocker

io.github.WYRE-AI由 GitHub 账号 WYRE-AI 发布v1.4.2更新于 Nov 21, 2026

MCP server for ThreatLocker — zero-trust endpoint protection, allowlisting, and policies.

概览

AI 生成的概览

将助手连接到 ThreatLocker 门户 API,用于查看计算机、审批请求、审计日志和组织,服务于零信任终端防护。

可以做什么
  • 导航各功能域并检查 API 连接状态
  • 列出并查看计算机详情、签到历史与计算机组
  • 列出、查看并统计审批请求,包括许可申请详情
  • 搜索审计日志条目、获取单条记录并查看文件审计历史
  • 列出子组织、获取组织的安装授权密钥,以及列出可用于移动计算机的组织
适用场景
适合已在使用 ThreatLocker 的安全或 IT 管理员,希望用自然语言查询终端清单、待处理审批和审计历史,而不必在门户中逐项点击。
运行要求
以 stdio 方式在本地运行(Docker 镜像或从源码构建的 Node.js 版本)。需要 ThreatLocker API 密钥(THREATLOCKER_API_KEY)和组织 ID(THREATLOCKER_ORGANIZATION_ID),并能访问 ThreatLocker API。网关/HTTP 模式还需 AUTH_MODE=gateway 与 CONDUIT_S2S_SECRET,并通过请求头逐次提供凭据。
安装前请注意
需要 ThreatLocker API 密钥这一敏感凭据;组织授权密钥工具会返回可用于注册的安装密钥。审批与审计工具仅在 API 用户角色具备相应门户权限时可用。网关模式通过请求头接收凭据,缺少有效 X-Gateway-S2S 证明的请求会被拒绝;请勿在非本地开发环境使用 MCP_ALLOW_INSECURE_DEV。

安装

在 SourceWeft 中

  1. 打开 控制台中的 ThreatLocker,将其添加到工作区。
  2. 为需要使用其工具的对话启用该服务。

Desktop only,通过 STDIO。 STDIO 服务会启动本地进程,因此需要 SourceWeft 桌面宿主。

其他 MCP 客户端

参照 仓库 中的启动说明。

README

ThreatLocker MCP Server

A Model Context Protocol (MCP) server that provides AI assistants with access to the ThreatLocker Portal API. Manage computers, approval requests, audit logs, and organizations through natural language interactions.

Features

  • Stateless Architecture: No session state required, fresh connections per request
  • Decision-Tree Navigation: Navigate domains with threatlocker_navigate
  • Gateway Mode: Multi-tenant support via HTTP headers
  • Elicitation Support: Interactive prompts for missing parameters
  • Comprehensive Error Handling: Detailed error messages and logging
  • Docker Support: Production-ready containerization

Tools

Navigation

  • threatlocker_navigate - Navigate to a domain to see available tools
  • threatlocker_status - Check API connection status and available domains

Computers

  • threatlocker_computers_list - List computers with filters (search, group, pagination)
  • threatlocker_computers_get - Get detailed computer information
  • threatlocker_computers_get_checkins - Get computer checkin history

Computer Groups

  • threatlocker_computer_groups_list - List computer groups with filters
  • threatlocker_computer_groups_dropdown - Get computer groups for dropdown selection

Approval Requests

  • threatlocker_approvals_list - List approval requests with status filters
  • threatlocker_approvals_get - Get detailed approval request information
  • threatlocker_approvals_pending_count - Get count of pending approvals
  • threatlocker_approvals_get_permit_application - Get permit application details

Troubleshooting: if approval tools return an HTTP 401 / "Authentication failed" while audit tools work with the same key, the ThreatLocker API user/role lacks Approval Center permissions. An admin must grant them in the ThreatLocker portal; retrying will not help.

Audit Log

  • threatlocker_audit_search - Search audit log entries with filters. searchText is a path filter (ThreatLocker fullPath), not free text: it matches the file path and also each row's process path and created-by path, so a program's path also finds what that program did. Give a full path or use * wildcards (*\setup.msi). Filter by computer with hostname. Pass childOrganizations: true for computers in child organizations. Filter by result with actionId (1 Permit, 2 Deny, 3 Deny with option to request, 6 Ringfenced, 99 Any Deny; onlyTrueDenies / simulateDeny split enforced from monitor-only denies) and by actionType (execute, install, network, ...). groupBys counts rows by one or two fields, so actionId: 2, groupBys: [6, 9] breaks denies down by policy name and action type. These filters need the SDK release that adds them; older SDKs send only actionType
  • threatlocker_audit_get - Get one audit log entry; pass eActionLogId, sourceTableId and organizationId straight from the threatlocker_audit_search row (organizationId is required for rows from child organizations; deprecated alias: actionLogId). HTTP 500 is reported as an actionable error: re-search and pass all three fields from the same row
  • threatlocker_audit_file_history - Get audit history for a file. Requires fullPath and either hostname or computerId. For a computer in a child organization also pass organizationId (from its threatlocker_computers_list row); without it the result is empty. Needs @wyre-ai/node-threatlocker with getFileHistory({ organizationId }) support

Organizations

  • threatlocker_organizations_list_children - List child organizations
  • threatlocker_organizations_get_auth_key - Get the installation auth key for the organization the configured credentials are scoped to. The result is a sensitive key and is never logged. Reports an explicit error instead of {} when ThreatLocker returns no JSON key, and returns text bodies with SDK 3.0.2
  • threatlocker_organizations_for_move_computers - Get organizations for computer moves

Configuration

Environment Variables

Stdio Mode (Direct API Access)
bash
THREATLOCKER_API_KEY=your_api_key_hereTHREATLOCKER_ORGANIZATION_ID=your_org_id_hereMCP_TRANSPORT=stdio
Gateway Mode (Multi-tenant)
bash
AUTH_MODE=gatewayMCP_TRANSPORT=httpMCP_HTTP_PORT=8080# Required. The HTTP server exits non-zero if this is empty.CONDUIT_S2S_SECRET=gateway-provisioned-secret# Optional. Unset binds 127.0.0.1. The container image sets 0.0.0.0.# MCP_HTTP_HOST=127.0.0.1

CONDUIT_S2S_SECRET is required for the HTTP transport. The process logs an error and exits non-zero when it is unset. For local development only, MCP_ALLOW_INSECURE_DEV=1 starts the server anyway and logs a loud warning; do not set that flag in production. The secret is never written to the logs.

When MCP_HTTP_HOST is unset, the server binds 127.0.0.1. The Dockerfile sets MCP_HTTP_HOST=0.0.0.0 so the process listens inside the container; publish that port on localhost.

Gateway Mode Headers

When running in gateway mode, include these headers with each request:

  • X-Gateway-S2S: Service-to-service proof from the gateway (t=<unix>,v1=<hmac>). Requests without a valid header are rejected with 401 when CONDUIT_S2S_SECRET is set.
  • X-Threatlocker-Api-Key: Your ThreatLocker API key (required). A request without it is rejected with 401 and does not fall back to THREATLOCKER_API_KEY.
  • X-Threatlocker-Organization-Id: Your organization ID (optional; the API defaults to the key's primary organization)

/health and /healthz stay unauthenticated and do not read credentials. The stdio transport does not use CONDUIT_S2S_SECRET.

Logging

bash
LOG_LEVEL=debug|info|warn|error  # Default: info

Local Development

  1. Clone the repository:
bash
git clone https://github.com/WYRE-AI/threatlocker-mcp.gitcd threatlocker-mcp
  1. Install dependencies:
bash
npm install
  1. Set environment variables:
bash
cp .env.example .env# Edit .env with your ThreatLocker credentials
  1. Build and run:
bash
npm run buildnpm start
# Or for development with hot reload:npm run dev
  1. Test the server:
bash
# Stdio modeecho '{"jsonrpc": "2.0", "id": 1, "method": "tools/list"}' | npm start
# HTTP mode. CONDUIT_S2S_SECRET is required unless you set# MCP_ALLOW_INSECURE_DEV=1 (local development only; logs a loud warning).# The server binds 127.0.0.1 unless MCP_HTTP_HOST is set.MCP_TRANSPORT=http AUTH_MODE=gateway CONDUIT_S2S_SECRET=dev-secret npm run start:httpcurl http://127.0.0.1:8080/health

Docker

Using Docker Compose

bash
# Required. Compose will not start if this is unset or empty.export CONDUIT_S2S_SECRET=gateway-provisioned-secret
# Pull and run latest image. The published port is 127.0.0.1:8080.docker compose up -d
# Or build locallydocker compose -f docker-compose.dev.yml up --build

Using Docker directly

bash
# Gateway mode (recommended). The container listens on 0.0.0.0 inside# the network namespace; publish it on localhost only. The process# exits if CONDUIT_S2S_SECRET is empty.docker run -d \  --name threatlocker-mcp \  -p 127.0.0.1:8080:8080 \  -e AUTH_MODE=gateway \  -e CONDUIT_S2S_SECRET=gateway-provisioned-secret \  ghcr.io/wyre-ai/threatlocker-mcp:latest
# Stdio modedocker run -d \  --name threatlocker-mcp \  -e THREATLOCKER_API_KEY=your_key \  -e THREATLOCKER_ORGANIZATION_ID=your_org_id \  -e MCP_TRANSPORT=stdio \  ghcr.io/wyre-ai/threatlocker-mcp:latest

Architecture

Directory Structure

src/├── domains/           # Domain-specific handlers│   ├── computers.ts│   ├── computer_groups.ts│   ├── approval_requests.ts│   ├── audit_log.ts│   ├── organizations.ts│   ├── navigation.ts│   └── index.ts├── utils/             # Utilities│   ├── client.ts      # ThreatLocker API client│   ├── logger.ts      # Structured logging│   ├── types.ts       # TypeScript types│   ├── server-ref.ts  # Server reference for elicitation│   └── elicitation.ts # Interactive prompts├── server.ts          # MCP server creation├── index.ts           # Stdio transport entry└── http.ts            # HTTP transport entry

Design Patterns

  • Domain Handlers: Each API area has its own handler with getTools() and handleCall()
  • Lazy Loading: Domain handlers are imported on-demand
  • Fresh Connections: New server instance per HTTP request for stateless operation
  • Credential Invalidation: Client is reset when credentials change
  • Elicitation Framework: Interactive prompts for missing parameters

License

Apache-2.0 - see LICENSE for details.

来源:README.md,提交 af389e2

工具

0
SourceWeft 尚未在沙箱中启动过该软件包,因此其工具未知。

版本历史

8
  1. v1.4.2最新Nov 21, 2026
  2. v1.4.1Nov 14, 2026
  3. v1.4.0Nov 12, 2026
  4. v1.3.14Nov 11, 2026
  5. v1.3.12Oct 10, 2026
  6. v1.3.11Oct 9, 2026
  7. v1.3.10Sep 24, 2026
  8. v1.3.9Sep 16, 2026

评分

已满足适用检查项的 40%

该评分衡量此收录条目的完整程度,依据注册中心条目、端点与公开来源计算得出。它不是对服务器本身的评价,也不表示服务器安全或当前可用:可连通性是带日期的检查结果,评分也不包含安装量或用户评价。

  • 近 30 天内可连通25 分

    请保持端点能够响应 MCP 握手;SourceWeft 每 30 天探测一次。

  • 已声明认证方式10 分

    请在注册中心条目中声明凭据请求头,或保持端点能够响应探测,以便判定其认证方式。

  • 已列出工具15 分

    请让服务器在无凭据时响应 tools/list,或发布可在沙箱中无凭据启动的软件包。

  • 描述必需15 分
  • 图标10 分
  • 文档15 分
  • 许可证10 分

    请在仓库或注册中心条目中声明许可证。