How mhtsec/artex works
ARTEX is an AI-driven autonomous penetration-testing system: users configure scope and a task, then the backend coordinates planner/worker agents to explore targets, record assets and findings, and expose progress and evidence in the web UI. The README also documents traffic capture, human approval/chat, vulnerability retesting, notifications, external asset synchronization, and MCP integrations. Edges cite the supplied excerpts or README; several substantial feature links are documented capabilities because their individual implementation files were not sampled.
Main entry points: server.go HTTP API and SSE · worker.go Worker agents · engine.go Exploration engine · db.go PostgreSQL domain store
Components 23
Web interface
- Task workspace
web/src/app/(main)/function/tasks/page.tsx - Findings views
web/src/app/(main)/function/findings/page.tsx - Human-in-loop chat
web/src/app/(main)/chat/page.tsx
Task control
- HTTP API and SSE
server/server.go - Task manager
server/manager.go - Exploration engine
server/engine.go
Agent exploration
- Planner
agent/planner.go - Worker agents
agent/worker.go - Security tools
agent/tools.go - Tool guard
guard/guard.go - Approval interception
intercept/intercept.go - Finding retester
agent/retester.go
Evidence and storage
- PostgreSQL domain store
db/db.go - Traffic recorder
traffic/traffic.go - Durable finding evidence
evidence/store.go - LLM call records
llmrec/llmrec.go - Finding reports
report/findings.go
External integrations
- LLM provider
agent/provider.go - MCP services
mcphttp/client.go - Notification channels
notify/notify.go - ScopeSentry
Other components
- Operator
- Authorized targets
Connections 27
- Operator → Task workspace: configures tasks
- Task workspace → HTTP API and SSE: calls API
- HTTP API and SSE → Task manager: manages lifecycle
server/server.go - HTTP API and SSE → Exploration engine: routes controls
server/server.go - Task manager → PostgreSQL domain store: persists tasks
server/manager.go - Exploration engine → Planner: runs planning
server/engine.go - Exploration engine → Worker agents: dispatches work
server/engine.go - Worker agents → Security tools: invokes tools
agent/worker.go - Worker agents → Tool guard: gates calls
agent/worker.go - Tool guard → Approval interception: requests review
guard/guard.go - Worker agents → PostgreSQL domain store: writes exploration
agent/worker.go - Worker agents → LLM provider: requests completion
- Exploration engine → LLM call records: records usage
- LLM call records → PostgreSQL domain store: stores call records
llmrec/llmrec.go - Authorized targets → Traffic recorder: exchanges HTTP
- Traffic recorder → PostgreSQL domain store: indexes exchanges
traffic/traffic.go - Durable finding evidence → Traffic recorder: snapshots exchanges
evidence/store.go - Durable finding evidence → PostgreSQL domain store: binds evidence
evidence/store.go - Findings views → HTTP API and SSE: requests findings
- HTTP API and SSE → Finding reports: generates exports
server/server.go - Operator → Human-in-loop chat: steers execution
- Human-in-loop chat → HTTP API and SSE: sends messages
- Exploration engine → Finding retester: runs retest agent
- HTTP API and SSE → Notification channels: dispatches notifications
- HTTP API and SSE → ScopeSentry: imports assets
- Worker agents → MCP services: uses MCP tools
- Traffic recorder → Authorized targets: records requests