$ omega status ============================================================ OMEGA Agent v0.4.0 — Channels ============================================================ [OK] Telegram: connected (@your_omega_bot) [OK] WhatsApp: ready [OK] Voice: Whisper (openai/whisper-1) active → voice messages transcribed before reaching provider [OK] Photos: inbox enabled → saved to ~/.omega/workspace/inbox/ → path injected into prompt for Claude Code Read tool → auto-cleaned after response [OK] Workspace: ~/.omega/workspace/ ============================================================ Non-blocking: each message spawns a concurrent task. Same-sender messages are serialized automatically. If a provider call is active, new messages are buffered with "Got it, I'll get to this next." and processed in order. ============================================================ Active conversations: 1 | Total messages: 2,847

$ omega status --providers ============================================================ OMEGA Agent — Provider Configuration ============================================================ Active: claude-code-cli → invokes: claude -p --output-format json → workspace: ~/.omega/workspace/ → auto-resume on max_turns (up to 5 attempts) → uses your local Claude auth — no extra API key Available providers: claude-code-cli ← current Claude Max ($200/mo flat) anthropic-api api.anthropic.com, pay per token openai api.openai.com, pay per token ollama stub localhost:11434, free (local GPU) openrouter stub proxy to 200+ models Switch in config.toml: [provider] name = "ollama" # free — runs entirely on your hardware Pre-flight planning: messages >15 words trigger a dedicated planning call first → DIRECT or step list. If steps returned → autonomous execution with progress updates + retry logic. ============================================================ Your keys. Your choice. Your cost.

[Ivan] buy 0.01 BTC futures, spread over 30 min → Algo Trading Skill — Binance Futures Strategy: TWAP (Time-Weighted Average Price) Symbol: BTCUSDT perpetual Size: 0.01 BTC (~$670 notional) Duration: 30 min → 6 child orders × 5 min intervals Risk check: Position size: 0.015% of equity ✓ (hard limit: 2%) Daily PnL: -0.1% ✓ (hard limit: -3%) Drawdown: 0.2% ✓ (circuit break: -5%) ⚠ TESTNET — confirm to execute? Reply YES to continue. [Ivan] yes [Order 1/6] BTCUSDT BUY 0.001667 @ $67,124.50 ✓ [Order 2/6] BTCUSDT BUY 0.001667 @ $67,089.20 ✓ [Order 3/6] BTCUSDT BUY 0.001667 @ $67,103.80 ✓ [Order 4/6] BTCUSDT BUY 0.001667 @ $67,118.60 ✓ [Order 5/6] BTCUSDT BUY 0.001667 @ $67,091.40 ✓ [Order 6/6] BTCUSDT BUY 0.001667 @ $67,099.70 ✓ Execution complete. Avg fill: $67,104.53 Slippage: 0.02% Duration: 30m 04s ──────────────────────────────────────────────────────────── Risk limits are hardcoded in Rust — the LLM cannot override. Human confirms every trade. Testnet by default.

User profile: name Ivan timezone America/Los_Angeles preferred_lang en active_project coding-assistant Recent context (last 5 messages): Ivan: what's the weather in SF today? OMEGA: It's 62°F and partly cloudy in San Francisco... Ivan: set a reminder for standup at 9am tomorrow OMEGA: Done! Reminder set for 9:00 AM. Ivan: can you summarize yesterday's git commits? Memory stats: Facts stored: 14 Summaries: 4 (auto-generated when history grows long) Scheduled tasks: 2 pending Audit entries: 891 Storage: ~/.omega/memory.db → SQLite, fully local, no sync, no cloud. → Facts extracted automatically from conversations. → Summaries replace raw history to keep context lean. → Scheduler polls every 60s and delivers due reminders.

$ omega status --sandbox ============================================================ OMEGA Agent — Sandbox ============================================================ Mode: sandbox Platform: macOS — Seatbelt (sandbox-exec) active ──────────────────────────────────────────────────────────── Write scope: ✓ ~/.omega/ (workspace, skills, projects, memory) ✓ /tmp (temporary files) ✓ ~/.claude (provider session state) ✗ /etc DENIED ✗ ~/Documents DENIED ✗ anywhere else DENIED ──────────────────────────────────────────────────────────── Isolation levels: sandbox OS-enforced + prompt rules ← current rx OS-enforced, reads unrestricted rwx Unrestricted (full trust mode) ──────────────────────────────────────────────────────────── Linux equivalent: Landlock LSM (kernel 5.13+) ============================================================ Your agent cannot write outside its data directory. No data leaves your machine without your explicit action. ============================================================