Security System

Key Features

• Per-tool permission controls (read, edit, bash, websearch, task)
• Per-agent permission profiles (Coder, Architect, Planner, Reviewer, Shell)
• Three permission levels: ALLOW (always permit), ASK (prompt user), DENY (block completely)
• Wildcard pattern support for flexible rule matching
• Configurable via .apex/config.yaml or CLI flags
• Permission checks happen before every tool execution

Configuration

# .apex/config.yaml
permissions:
  default: ask
  rules:
    - tool: read_file
      action: allow
    - tool: run_command
      action: ask
      patterns:
        - "rm *"     # Always ask for rm
        - "sudo *"   # Always ask for sudo
    - tool: delete_file
      action: deny  # Never allow file deletion

Key Features

• Configurable rate limits per model and per session
• Memory backend for fast, in-process rate tracking
• SQLite backend for persistent rate limiting across sessions
• Token-based and request-based limiting
• Automatic backoff and retry logic
• Cost threshold alerts and budget limits

Configuration

# .apex/config.yaml
rate_limiting:
  enabled: true
  backend: sqlite  # or "memory"
  limits:
    - model: "anthropic/*"
      requests_per_minute: 60
      tokens_per_minute: 100000
    - model: "openai/*"
      requests_per_minute: 100
  budget:
    daily_limit: 10.00    # $10/day
    monthly_limit: 200.00 # $200/month

Key Features

• Command analysis and classification before execution
• Pattern-based dangerous command detection (rm -rf, sudo, chmod 777)
• Blocklist of known dangerous commands
• Allowlist for trusted command patterns
• Shell injection prevention
• Command audit logging for compliance

Configuration

# .apex/config.yaml
shell_security:
  enabled: true
  block_patterns:
    - "rm -rf /"
    - "dd if=* of=/dev/*"
    - ":(){ :|:& };:"
    - "chmod 777"
    - "> /dev/sda"
  ask_patterns:
    - "sudo *"
    - "pip install *"
    - "npm install -g *"
  allow_patterns:
    - "git *"
    - "pytest *"
    - "ls *"

Key Features

• Workspace-based key storage with SHA-256 hashing
• Keys never appear in logs or conversation history
• Automatic key rotation support
• Key validation on startup
• Support for multiple keys per provider (load balancing)
• Key usage tracking and audit log

Configuration

# ~/.apex/.env (encrypted at rest)
ANTHROPIC_API_KEY=sk-ant-...
OPENAI_API_KEY=sk-...

# Key rotation
apex /key rotate anthropic
apex /key validate
apex /key status

Key Features

• All file paths resolved against the project root
• Symlink resolution and validation
• Blocking of ../ traversal patterns
• Configurable allowed and denied paths
• Protection against absolute path injection
• Workspace boundary enforcement

Configuration

# .apex/config.yaml
path_security:
  enabled: true
  allowed_paths:
    - "."           # Project root
    - "/tmp/apex"   # Temp directory
  denied_paths:
    - "/etc"
    - "/root"
    - "~/.ssh"
    - "~/.gnupg"
  resolve_symlinks: true
  max_path_length: 4096

Key Features

• Model-specific pricing with accurate per-token cost calculation
• Real-time cost display in TUI and CLI
• Session cost tracking with /cost command
• Three billing plans: Free, Pro ($200/mo), Enterprise (custom)
• Daily and monthly budget limits with alerts
• Cost breakdown by model and provider
• Export cost reports as CSV or JSON

Configuration

# Check costs
apex /cost

# Cost output:
# Session:  $0.42 (1,234 tokens)
# Today:    $3.21
# This month: $47.89 / $200.00 budget

# Budget configuration
budget:
  daily: 10.00
  monthly: 200.00
  alert_threshold: 0.8  # Alert at 80%

Key Features

• Isolated execution environments for untrusted code
• Support for Python, JavaScript, Bash, Ruby, Go, Rust
• Configurable timeout (default: 30 seconds)
• Memory and CPU limits
• Network isolation by default
• Output capture and return

Configuration

# Code runs in sandbox
apex "Run this Python code: print('hello')"

# Sandbox config
sandbox:
  enabled: true
  timeout: 30
  max_memory_mb: 512
  max_output_chars: 10000
  network: disabled

Key Features

• Bearer token or X-API-Key authentication
• Per-endpoint rate limiting
• Automatic cost tracking per request
• Shell security integration for API-initiated commands
• CORS and origin validation
• Request logging and audit trail

Configuration

# Start secure HTTP API server
from apex.http_api import HTTPServer

server = HTTPServer(
    host="127.0.0.1",
    port=8080,
    require_auth=True,
)

# API key required for all requests
# Authorization: Bearer <api-key>
# or X-API-Key: <api-key>