🇪🇸 Leer en español

SiriusOS

Persistent 24/7 Claude Code agents you control from Telegram or your phone.

Telegram chat

You:     Morning. What did you ship overnight?
Boss:    Overnight recap: completed 4 tasks, ran 2 theta-wave
         experiments, drafted 3 content scripts. One item needs
         your approval — I want to email the beta waitlist.
         Check the dashboard or reply "approve".

You:     approve
Boss:    Sent. Email delivered to 47 recipients. Task closed.

You:     Add a cron to check my inbox every morning at 8am.
Boss:    Done. "morning-inbox" cron set — runs daily at 08:00.
         Saved to crons.json — survives restarts automatically.

Features

• Persistent agents — Claude Code runs 24/7 in PTY sessions, auto-restarting on crash or after 71-hour context rotation.
• Multi-agent orchestration — Orchestrator, Analyst, and specialist agents coordinate via a shared file bus. Tasks, blockers, and approvals flow automatically.
• Telegram + iOS control — Send commands, approve actions, and get reports from anywhere. Native iOS app coming soon.
• Web dashboard — Full-featured Next.js UI for tasks, approvals, experiments, analytics, and agent fleet health.
• Autoresearch (theta wave) — Agents run autonomous experiments overnight, evaluate results, and surface findings for your review.

Architecture

flowchart TD
    U["User (Telegram / iOS)"] --> CLI["SiriusOS Daemon (Node.js)"]
    CLI --> O["Orchestrator agent"]
    CLI --> A["Analyst agent"]
    CLI --> W["Specialist agents"]
    O <-->|file bus| A
    O <-->|file bus| W
    CLI --> D["Web Dashboard (Next.js)"]
    D --> U2["Browser / iOS App"]

Quick Start

Requirements: Node.js 20+, Claude API key, Telegram bot token from @BotFather.

Option A — One-line installer (recommended)

macOS / Linux:

curl -fsSL https://siriusos.unikprompt.com/install.mjs | node

Windows (PowerShell, requires WSL2):

node -e "$(irm https://siriusos.unikprompt.com/install.mjs)"

The Node-based installer clones the repo into ~/siriusos, verifies prerequisites (Node 20+, jq, claude CLI, build tools), and on Windows checks for WSL2 — agents run shell scripts under bash, so WSL is required (the installer points you to wsl --install if missing).

Override the install location with SIRIUSOS_DIR=/custom/path or pin a branch with SIRIUSOS_BRANCH=feature/foo.

Option B — Visual wizard from the dashboard

npm install -g pm2
npm install -g siriusos
siriusos dashboard --build

Open http://localhost:3013, sign in with the credentials from ~/.siriusos/default/dashboard.env, and you are routed automatically to /onboarding if no organization exists yet.

Option C — Manual (advanced)

npm install -g pm2
npm install -g siriusos

siriusos install
siriusos init myorg
siriusos add-agent boss --template orchestrator --org myorg
siriusos add-agent analyst --template analyst --org myorg

cat > orgs/myorg/agents/boss/.env <<EOF
BOT_TOKEN=<your-bot-token>
CHAT_ID=<your-chat-id>
ALLOWED_USER=<your-telegram-user-id>
EOF

siriusos ecosystem
pm2 start ecosystem.config.js && pm2 save && pm2 startup

# Windows: pm2 startup is unsupported. Use Task Scheduler instead:
#   powershell -ExecutionPolicy Bypass -File scripts\install-windows-pm2-startup.ps1

The orchestrator comes online in Telegram and you finish setup from there.

Open the dashboard later

siriusos dashboard --build --port 3013
cat ~/.siriusos/default/dashboard.env  # credentials

The ES|EN toggle lives in the navbar and in Settings → Appearance → Language.

Requirements

| Dependency | Notes | |---|---| | Node.js 20+ | nodejs.org | | macOS, Linux, or Windows 10/11 | Windows uses Task Scheduler for reboot persistence — see scripts/install-windows-pm2-startup.ps1 | | Claude Code | npm install -g @anthropic-ai/claude-code + claude login | | PM2 | npm install -g pm2 | | Telegram bot token | Create via @BotFather |

Voice Transcription (optional)

Telegram voice messages can be auto-transcribed locally with whisper.cpp before they reach the agent. SiriusOS invokes the local whisper-cli binary with a GGML model, so no hosted API is required. Install once to enable:

brew install whisper-cpp ffmpeg
bash scripts/install-whisper-model.sh

Zero config: defaults are model ggml-base.bin at ~/.siriusos/models/, language es, and whisper-cli/ffmpeg from PATH. Override with CTX_WHISPER_MODEL, CTX_WHISPER_LANG, CTX_WHISPER_BIN, or CTX_FFMPEG_BIN. If local transcription is missing or fails, the daemon falls back to the existing local_file: injection so the agent can still reach the audio.

Templates

| Template | Description | |---|---| | orchestrator | Coordinates agents, manages goals, handles morning/evening reviews, approves actions | | analyst | System health, metrics, theta-wave autoresearch, analytics | | agent | General-purpose worker — use this as the base for specialist agents |

CLI Reference

siriusos install            # Set up state directories
siriusos init <org>         # Create an organization
siriusos add-agent <name>   # Add an agent (--template, --org)
siriusos enable <name>      # Enable agent in daemon
siriusos ecosystem          # Generate PM2 config
siriusos status             # Agent health table
siriusos doctor             # Check prerequisites
siriusos list-agents        # List agents
siriusos dashboard          # Start web dashboard (--port 3000)

Security

SiriusOS has undergone a dedicated security hardening sprint covering prompt injection resistance, guardrail enforcement, and approval gate integrity. Agents require explicit human approval before any external action (email, deploy, delete, financial). The guardrails system is self-improving: agents log near-misses and extend GUARDRAILS.md each session.

License

MIT — see LICENSE.