Sleepless Agent

A 24/7 AgentOS that works while you sleep

Have Claude Code Pro but not using it at night? Transform it into an AgentOS that handles your ideas and tasks while you sleep. This is a 24/7 AI assistant daemon powered by Claude Code CLI and Python Agent SDK that processes both random thoughts and serious tasks via Slack with isolated workspaces.

📰 News

• [2025-10-26] 🎉 Initial release v0.1.0 - Full AgentOS with multi-agent workflow support
• [2025-10-25] 🚀 Added task auto-generation with configurable strategies
• [2025-10-24] 🔧 Integrated Git management with automatic PR creation
• [2025-10-23] 📊 Implemented isolated workspaces for parallel task execution
• [2025-10-22] 💡 Added Claude Code Python Agent SDK integration

🎬 Demo

Quick Example

# Start the daemon
$ sle daemon
2025-10-26 03:30:12 | INFO | Sleepless Agent starting...
2025-10-26 03:30:12 | INFO | Slack bot connected

# Submit a task via Slack
/think Implement OAuth2 authentication -p backend

# Check status
$ sle check
╭─────────────────── System Status ───────────────────╮
│ 🟢 Daemon: Running                                  │
│ 📊 Queue: 3 pending, 1 in_progress                  │
│ 💻 Usage: 45% (Day threshold: 95%)                  │
│ 🔄 Last task: "Implement OAuth2..." (in progress)   │
╰─────────────────────────────────────────────────────╯

# View results
$ sle report 42
Task #42: ✅ Completed
Branch: feature/backend-42
PR: https://github.com/user/repo/pull/123

✨ Features

• 🤖 Continuous Operation: Runs 24/7 daemon, always ready for new tasks
• 💬 Slack Integration: Submit tasks via Slack commands
• 💭 Interactive Chat Mode: Real-time conversational sessions with Claude in Slack threads
• 🎯 Hybrid Autonomy: Auto-applies random thoughts, requires review for serious tasks
• ⚡ Smart Scheduling: Optimizes task execution based on priorities
• 📊 Task Queue: SQLite-backed persistent task management
• 🔌 Claude Code SDK: Uses Python Agent SDK to interface with Claude Code CLI
• 🏗️ Isolated Workspaces: Each task gets its own workspace for true parallelism
• 📝 Result Storage: All outputs saved with metadata for future reference

⚙️ Prerequisites

• Python 3.11+
• Claude Code CLI installed (npm install -g @anthropic-ai/claude-code)
• Git (for auto-commits)
• gh CLI (optional, for PR automation)
• Slack workspace admin access (optional — the sle CLI works without Slack)

💡 Slack is optional. You can run all sle CLI commands without configuring Slack. Slack integration adds a convenient real-time interface for submitting tasks and checking status, but every feature is also available via the CLI.

🚀 Quick Start

1. Install

pip install sleepless-agent

Or for development:

git clone <repo>
cd sleepless-agent
python -m venv venv
source venv/bin/activate  # or venv\Scripts\activate on Windows
pip install -e .

⚠️ Windows / WSL tip: After pip install, the sle command might not be found if Python's Scripts directory is not on your PATH. See Troubleshooting: sle not found below.

2. Authenticate Claude Code (Required)

# Install the Claude Code CLI (requires Node.js)
npm install -g @anthropic-ai/claude-code

# Log in once (opens a browser)
claude login

# Verify
claude --version

3. (Optional) Start without Slack

Slack integration is not required. You can use the sle CLI directly:

# Start the daemon (no Slack needed)
sle daemon

# Queue a task
sle think "Research async Python patterns"

# Check status
sle check

If you want Slack integration, continue with the setup below.

4. Setup Slack App

Visit https://api.slack.com/apps and create a new app:

Basic Information

• Choose "From scratch"
• Name: "Sleepless Agent"
• Pick your workspace

Enable Socket Mode

• Settings > Socket Mode > Toggle ON
• Generate app token (starts with xapp-)

Create Slash Commands Settings > Slash Commands > Create New Command:

• /think - Capture thought or task (use -p project-name for serious tasks)
• /chat - Start interactive chat mode with Claude
• /check - Check queue status
• /usage - Show Claude Code Pro plan usage
• /cancel - Cancel task or project
• /report - Show reports or task details
• /trash - Manage trash (list, restore, empty)

OAuth Scopes Features > OAuth & Permissions > Bot Token Scopes:

• chat:write
• commands
• app_mentions:read
• channels:history (for chat mode)
• groups:history (for chat mode in private channels)
• reactions:write (for chat mode indicators)

Event Subscriptions (for Chat Mode) Features > Event Subscriptions > Enable Events > Subscribe to bot events:

• message.channels
• message.groups

Install App

• Install to workspace
• Get bot token (starts with xoxb-)

5. Configure Environment

cp .env.example .env
nano .env  # Edit with your tokens

Set:

• SLACK_BOT_TOKEN - xoxb-... token
• SLACK_APP_TOKEN - xapp-... token

(Claude API key no longer needed - uses Claude Code CLI)

6. Run

sle daemon

You should see startup logs similar to:

2025-10-24 23:30:12 | INFO     | sleepless_agent.interfaces.bot.start:50 Slack bot started and listening for events
2025-10-24 23:30:12 | INFO     | sleepless_agent.runtime.daemon.run:178 Sleepless Agent starting...

Logs are rendered with Rich for readability; set SLEEPLESS_LOG_LEVEL=DEBUG to increase verbosity.

🖥️ Windows / WSL: sle Command Not Found

If sle is not recognised after pip install on Windows or WSL, the Python Scripts directory is likely not on your PATH.

Windows (PowerShell / CMD)

# Find where pip installed the script
python -m site --user-scripts

# Add that directory to PATH permanently (PowerShell)
$scriptsDir = python -m site --user-scripts
[Environment]::SetEnvironmentVariable("PATH", "$env:PATH;$scriptsDir", "User")

# Or run sle directly with python -m
python -m sleepless_agent.interfaces.cli --help

WSL (Ubuntu / Debian)

# Add the local bin directory to PATH
echo 'export PATH="$HOME/.local/bin:$PATH"' >> ~/.bashrc
source ~/.bashrc

# Verify
sle --version

System-wide pip install (not recommended)

If you used sudo pip install, the script will be in /usr/local/bin which is normally already in PATH. However, prefer using a virtual environment or pip install --user.

Virtual environment (recommended)

Using a virtual environment avoids PATH issues on all platforms:

python -m venv venv
# Windows:
venv\Scripts\activate
# macOS / Linux / WSL:
source venv/bin/activate

pip install sleepless-agent
sle --version  # always works inside the venv

💬 Slack Commands

All Slack commands align with the CLI commands for consistency:

📋 Task Management

| Command | Purpose | Example | |---------|---------|---------| | /think | Capture random thought | /think Explore async ideas | | /think -p <project> | Add serious task to project | /think Add OAuth2 support -p backend | | /check | Show system status | /check | | /usage | Show Claude Code Pro usage | /usage | | /cancel | Cancel task or project | /cancel 5 or /cancel my-app |

💭 Interactive Chat Mode

Start a real-time conversation with Claude in a dedicated Slack thread:

| Command | Purpose | Example | |---------|---------|---------| | /chat <project> | Start chat mode for a project | /chat my-backend | | /chat end | End current chat session | /chat end | | /chat status | Check active session status | /chat status | | /chat help | Show chat mode help | /chat help |

Chat Mode Features:

• 🧵 Dedicated thread for each session
• 💬 Full conversation history maintained
• 🔄 Real-time processing indicator
• 📁 Claude can read/write/edit files in project workspace
• ⏱️ Auto-timeout after 30 minutes of inactivity
• Type exit in thread to end session

💡 Note: When you run /chat <project>, a new thread is created. All your prompts must be sent inside this thread - Claude only responds to messages within the chat thread, not in the main channel.

📊 Reporting & Trash

| Command | Purpose | Example | |---------|---------|---------| | /report | Today's report, task details, date/project report, or list all | /report, /report 42, /report 2025-10-22, /report my-app, /report --list | | /trash | List, restore, or empty trash | /trash list, /trash restore my-app, /trash empty |

⌨️ Command Line Interface

Install the project (or run within the repo) and use the bundled CLI:

python -m sleepless_agent.interfaces.cli think "Ship release checklist" -p my-app
# or, after installing the package:
sle check

The CLI mirrors the Slack slash commands:

| Command | Purpose | Example | |---------|---------|-