proffesor-for-testing__agentic-qe/docs/cli/agent-commands.md

Agent CLI Commands Reference

Complete reference for all 10+ agent management CLI commands in the AQE Fleet system.

Overview

The AQE Fleet provides comprehensive agent management through CLI commands organized into four categories:

1. Lifecycle Commands - Spawn, kill, restart agents
2. Monitoring Commands - List, metrics, logs, inspect
3. Task Commands - Assign tasks to agents
4. Interactive Commands - Attach/detach console sessions

---

Lifecycle Commands

agent spawn

Spawn a new QE agent with the specified type and task.

Usage:

aqe agent spawn <agentType> --task "<task description>" [options]

Agent Types:

Type	Description
test-generator	AI-powered test generation
coverage-analyzer	Coverage gap detection (O(log n))
quality-gate	Quality gate decisions
performance-tester	Performance and load testing
security-scanner	Security vulnerability detection
flaky-test-hunter	Flaky test detection and fixes
chaos-engineer	Chaos engineering experiments
visual-tester	Visual regression testing
code-reviewer	Code quality review
integration-tester	Integration test execution

Options:

• --task <description> - Task for the agent to execute (required)
• --model <model> - Model to use: claude-sonnet, claude-opus (default: claude-sonnet)
• --max-tokens <n> - Maximum tokens (default: 8192)
• --wait - Wait for completion and show output
• --json - Output in JSON format

Examples:

# Generate tests for a service
aqe agent spawn test-generator --task "Create unit tests for UserService"

# Analyze coverage gaps
aqe agent spawn coverage-analyzer --task "Find untested code paths in auth module"

# Security scan with opus model
aqe agent spawn security-scanner --task "Audit API endpoints" --model claude-opus

# Wait for completion
aqe agent spawn test-generator --task "Generate tests" --wait

Output:

{
  "success": true,
  "agentId": "agent-550e8400-e29b-41d4-a716-446655440000",
  "agentType": "test-generator",
  "status": "running",
  "startedAt": "2025-01-03T10:30:00.000Z"
}

---

Alternative: Edge Server REST API

For external integrations, CI/CD pipelines, or web applications, you can spawn agents via the Edge Server REST API instead of CLI.

Start Edge Server:

# Start combined HTTP + WebSocket server
node dist/edge/server/index.js

# Or with custom ports
HTTP_PORT=3001 WS_PORT=3002 node dist/edge/server/index.js

REST API Endpoints:

Endpoint	Method	Description
POST /api/agents/spawn	POST	Spawn new agent
GET /api/agents	GET	List all agents
GET /api/agents/:id	GET	Get agent status
GET /api/agents/:id/output	GET	Get agent output
DELETE /api/agents/:id	DELETE	Cancel agent
GET /api/agents/types	GET	List available types
GET /api/health	GET	Health check

Spawn Agent via REST:

curl -X POST http://localhost:3001/api/agents/spawn \
  -H "Content-Type: application/json" \
  -d '{
    "agentType": "test-generator",
    "task": "Create unit tests for UserService",
    "model": "claude-sonnet-4-20250514"
  }'

Response:

{
  "success": true,
  "agentId": "agent-550e8400-e29b-41d4-a716-446655440000",
  "status": "running"
}

When to Use:

• CLI (aqe agent spawn): Direct terminal usage, scripting
• REST API: CI/CD integration, web apps, external tools

---

agent restart

Gracefully restart an agent while preserving configuration and state.

Usage:

aqe agent restart <agentId> [options]

Options:

• --preserve-state - Save and restore agent state (default: true)
• --timeout <ms> - Termination timeout in milliseconds (default: 30000)
• --force - Force restart without graceful shutdown
• --no-state - Skip state preservation

Examples:

# Graceful restart with state preservation
aqe agent restart agent-1 --preserve-state

# Force restart
aqe agent restart agent-1 --force

# Custom timeout
aqe agent restart agent-1 --timeout 60000

Output:

{
  "agentId": "agent-1",
  "oldInstanceId": "test-generator-1-...",
  "newInstanceId": "test-generator-2-...",
  "status": "active",
  "restartTime": 1534,
  "stateRestored": true
}

---

Monitoring Commands

agent inspect

Display detailed agent information including configuration, metrics, history, and health status.

Usage:

aqe agent inspect <agentId> [options]

Options:

• --history - Include task execution history
• --metrics - Include performance metrics (default: true)
• --logs - Include recent log entries
• --format <type> - Output format: json, yaml, table, detailed (default: json)
• --depth <level> - Inspection depth: shallow, deep (default: deep)

Examples:

# Basic inspection
aqe agent inspect agent-1

# Full inspection with history and logs
aqe agent inspect agent-1 --history --logs

# Table format for terminal viewing
aqe agent inspect agent-1 --format table

# YAML output
aqe agent inspect agent-1 --format yaml

Output:

{
  "id": "test-generator-1-...",
  "type": "TEST_GENERATOR",
  "mcpType": "test-generator",
  "status": "active",
  "configuration": {
    "name": "agent-1",
    "capabilities": ["property-testing", "mutation-testing"],
    "resources": {
      "cpu": "1",
      "memory": "512MB"
    }
  },
  "lifecycle": {
    "spawnedAt": "2025-10-06T13:30:00.000Z",
    "lastActivity": "2025-10-06T13:37:00.000Z",
    "uptime": 420000,
    "restartCount": 2
  },
  "metrics": {
    "tasksCompleted": 47,
    "tasksActive": 1,
    "tasksFailed": 3,
    "averageExecutionTime": 1523,
    "successRate": 0.936
  },
  "health": {
    "status": "healthy",
    "issues": [],
    "lastCheck": "2025-10-06T13:37:10.000Z"
  }
}

---

Task Commands

agent assign

Assign tasks to agents with intelligent load balancing and capability matching.

Usage:

aqe agent assign <taskId> [options]

Options:

• --agent <agentId> - Manually assign to specific agent
• --auto-balance - Use load balancing algorithm
• --require-capability <name> - Require specific capability
• --priority <level> - Task priority: low, medium, high, critical (default: medium)
• --timeout <ms> - Task timeout in milliseconds (default: 300000)
• --retry - Retry on failure (default: true)

Examples:

# Manual assignment
aqe agent assign task-123 --agent agent-1

# Auto-assignment with load balancing
aqe agent assign task-123 --auto-balance

# Require specific capability
aqe agent assign task-123 --auto-balance --require-capability performance-testing

# High priority task
aqe agent assign task-123 --agent agent-1 --priority high

Output:

{
  "taskId": "task-123",
  "agentId": "agent-1",
  "agentType": "performance-tester",
  "status": "assigned",
  "assignedAt": "2025-10-06T13:37:15.000Z",
  "estimatedCompletion": "2025-10-06T13:39:00.000Z",
  "capabilities": ["load-testing", "stress-testing"]
}

Queued Response:

{
  "taskId": "task-124",
  "agentId": "agent-1",
  "agentType": "performance-tester",
  "status": "queued",
  "assignedAt": "2025-10-06T13:37:20.000Z",
  "queuePosition": 2,
  "capabilities": ["load-testing"]
}

---

Interactive Commands

agent attach

Attach to an agent's console for real-time monitoring of logs, metrics, and events.

Usage:

aqe agent attach <agentId> [options]

Options:

• --follow - Follow log output in real-time (default: true)
• --metrics - Show metrics updates (default: true)
• --logs - Show log entries (default: true)
• --events - Show event notifications (default: true)
• --refresh <ms> - Metrics refresh rate in milliseconds (default: 1000)
• --filter <pattern> - Filter logs by pattern

Examples:

# Attach with all monitoring
aqe agent attach agent-1

# Attach with only logs
aqe agent attach agent-1 --no-metrics --no-events

# Filter error logs
aqe agent attach agent-1 --filter ERROR

# Custom refresh rate
aqe agent attach agent-1 --refresh 500

Output:

============================================================
📎 Attached to Agent: test-generator-1-...
============================================================
Type: test-generator
Status: active
Tasks Completed: 47
Last Activity: 10/6/2025, 1:37:00 PM
============================================================

[LOG] Starting test generation for module: UserService
[METRIC] {"tasksCompleted":48,"averageExecutionTime":1498}
[EVENT] task_started: Generating unit tests
[LOG] Generated 15 test cases
[LOG] Running property-based testing...
[METRIC] {"tasksCompleted":48,"averageExecutionTime":1501}
[EVENT] task_completed: Test generation completed

Detach: Press Ctrl+C or use aqe agent detach <agentId>

---

agent detach

Detach from an active agent console session with optional session archival.

Usage:

aqe agent detach <agentId> [options]

Options:

• --save-session - Archive session data (default: true)
• --show-stats - Display session statistics (default: true)
• --force - Force detach even if errors occur

Examples:

# Normal detach with stats
aqe agent detach agent-1

# Force detach without saving
aqe agent detach agent-1 --force --no-save-session

Output:

============================================================
📊 Session Statistics
============================================================
Agent ID: agent-1
Session ID: attach-agent-1-1728223030000
Duration: 124.53s
Logs Received: 342
Events Received: 28
Metrics Received: 124
============================================================

---

Advanced Usage

Load Balancing Algorithm

The agent assign --auto-balance command uses a sophisticated scoring algorithm:

score = 100
  - (tasksCompleted * 0.1)        // Penalize by load
  - (status === 'busy' ? 50 : 0)  // Penalize busy agents
  - (status === 'error' ? 80 : 0) // Penalize error agents
  + (status === 'idle' ? 20 : 0)  // Bonus for idle
  + max(0, 20 - (avgTime / 1000)) // Bonus for fast agents

Health Checks

Agent health is automatically monitored:

• Healthy: Normal operation, no issues
• Degraded: Inactive for >1 hour, or no tasks completed after 10min
• Unhealthy: Agent in error state

Session Management

Attach sessions are automatically managed:

• Session metadata stored in .aqe/sessions/
• Completed sessions archived in .aqe/sessions/archive/
• Auto-cleanup of sessions older than 30 days

---

Integration with AgentRegistry

All commands integrate with the centralized AgentRegistry service:

import { getAgentRegistry } from '@/mcp/services/AgentRegistry';

const registry = getAgentRegistry();
const agent = registry.getRegisteredAgent(agentId);

---

Error Handling

Common Errors

Agent Not Found:

Error: Agent not found: agent-xyz

Solution: Check agent ID with aqe agent list

Agent Lacks Capability:

Error: Agent lacks required capability: performance-testing

Solution: Use aqe agent inspect <agentId> to check capabilities

Not Attached:

Error: Not attached to agent: agent-1

Solution: Use aqe agent attach agent-1 first

---

File Locations

• Agent configs: .aqe/agents/<agentId>.json
• Agent state: .aqe/state/<agentId>.state.json
• Agent logs: .aqe/logs/<agentId>.log
• Task queue: .aqe/queue/<agentId>.json
• Attach sessions: .aqe/sessions/<sessionId>.json
• Session archive: .aqe/sessions/archive/<sessionId>.json

---

Testing

All commands have comprehensive test coverage:

npm test -- tests/cli/agent.test.ts

Test Suites:

• ✅ agent restart (6 tests)
• ✅ agent inspect (7 tests)
• ✅ agent assign (7 tests)
• ✅ agent attach (6 tests)
• ✅ agent detach (7 tests)

Total: 33 tests covering all command functionality

---

Related Commands

• aqe fleet init - Initialize AQE fleet
• aqe fleet status - Show fleet overview
• aqe agent spawn - Create new agent
• aqe agent list - List all agents
• aqe agent metrics - Show agent metrics

---

Support

For issues or questions:

• GitHub: https://github.com/ruvnet/agentic-qe
• Documentation: /docs