convez.ai
MCP ServerSegurançaAvançado
wyre-technology

ThreatLocker MCP Server

wyre-technology/threatlocker-mcp

Integração ThreatLocker: allowlisting, aprovações e auditoria via MCP

Ver no GitHub
InicialPluginScore · de 0 a 100

Servidor MCP que integra assistentes de IA com a plataforma ThreatLocker para gerenciamento de zero-trust application allowlisting. Oferece acesso via linguagem natural a recursos de computers, approval requests, audit logs e organizações através da API do ThreatLocker Portal.

Instalação

$npx convez-ai wyre-technology/threatlocker-mcp

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
1
Forks
6
Acompanhando
0
Saúde
62%
Último commit
há 3 dias
Versão
v1.4.1
Issues abertas
6
Contribuidores
7

Casos de uso

  • Consultar e gerenciar solicitações de aprovação de aplicações em zero-trust
  • Buscar e analisar logs de auditoria com filtros avançados (caminho, política, ação)
  • Listar e obter detalhes de computers e grupos em ambientes multi-tenant

Veredito

MCP robusto e bem-estruturado para integração com ThreatLocker, voltado a segurança zero-trust em nível empresarial. Requer conhecimento prévio da plataforma ThreatLocker e permissões adequadas de API; implementação pronta para produção com tratamento completo de erros e suporte multi-tenant.

README

A tradução para o português deste README ainda está na fila.

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)

THREATLOCKER_API_KEY=your_api_key_here
THREATLOCKER_ORGANIZATION_ID=your_org_id_here
MCP_TRANSPORT=stdio

Gateway Mode (Multi-tenant)

AUTH_MODE=gateway
MCP_TRANSPORT=http
MCP_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

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

Local Development

  1. Clone the repository:
git clone https://github.com/WYRE-AI/threatlocker-mcp.git
cd threatlocker-mcp
  1. Install dependencies:
npm install
  1. Set environment variables:
cp .env.example .env
# Edit .env with your ThreatLocker credentials
  1. Build and run:
npm run build
npm start

# Or for development with hot reload:
npm run dev
  1. Test the server:
# Stdio mode
echo '{"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:http
curl http://127.0.0.1:8080/health

Docker

Using Docker Compose

# 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 locally
docker compose -f docker-compose.dev.yml up --build

Using Docker directly

# 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 mode
docker 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.

Ver o original no GitHub

Bom para

Relacionados