maxs 0.27.0

minimalist ai agent

maxs

Build AI tools while talking to them

Stop writing configs. Stop restarting servers. Just speak:

"max, create a tool that monitors our Kubernetes cluster"
→ Tool created, tested, and running in 30 seconds

🚀 Try it now (2 minutes)

pipx install maxs && maxs

Then just speak:

• "max, what time is it?" → Instant response
• "max, create a weather tool" → Tool appears in ./tools/
• "max, use the weather tool for Tokyo" → Working immediately

No configuration. No API keys. No YAML files.

💡 How it works

Voice Command → Hot Reload Tool Creation → Instant Usage
     ↓                    ↓                    ↓
"max, create X"    ./tools/X.py saved    Tool ready to use

Three revolutionary features:

🎙️ Voice-First Development - Build tools by talking
🔥 Instant Hot Reload - Tools work immediately, no restart
📡 P2P Agent Network - Agents communicate over encrypted mesh

⚡ Setup (choose one)

Option 1: Local AI (recommended for privacy)

curl -fsSL https://ollama.ai/install.sh | sh
ollama pull qwen3:4b
maxs

Option 2: Cloud AI (recommended for power)

# Anthropic Claude (best for development)
export ANTHROPIC_API_KEY="your-key"
MODEL_PROVIDER=anthropic maxs

# OpenAI GPT (good for general tasks)  
export OPENAI_API_KEY="your-key"
MODEL_PROVIDER=openai maxs

# Other: bedrock, github, litellm, llamaapi, mistral

🔥 Key Features

🎙️ Voice-Powered Development

maxs
> listen(action="start", trigger_keyword="max")

# Then just speak naturally:
"max, analyze the server logs"     → Creates monitoring tools
"max, deploy to staging"          → Runs deployment scripts  
"max, create a backup system"     → Builds automation instantly

🔄 Instant Hot Reload

Create tools while maxs is running:

# Save this to ./tools/weather.py
from strands import tool

@tool
def weather(city: str) -> str:
    return f"Weather for {city}: 72°F, sunny"

Then immediately use it:

maxs "get weather for Tokyo"  # Tool works instantly!

📡 P2P Agent Network

Agents communicate over encrypted Bluetooth mesh:

# Agent Alice says:
"max, analyze the server performance"

# Agent Bob automatically responds:
"Performance analysis complete. CPU at 85%, recommend scaling."

# No internet required - direct agent-to-agent communication

🧠 Intelligent Memory

Remembers everything across sessions:

maxs "remember: we're using PostgreSQL for user data"
# Later...
maxs "create a user analytics tool"  # Uses PostgreSQL context

🛠️ Common Workflows

Development & Automation

maxs "analyze this codebase and suggest improvements"
maxs "create deployment script for our Docker app"  
maxs "monitor system logs in the background"
maxs "format all Python files in this project"

Data Analysis

maxs "connect to PostgreSQL and analyze user growth"
maxs "create charts from the sales CSV file"
maxs "query the API and generate a report"

Voice Control (Hands-Free)

# Start voice mode first
maxs
> listen(action="start", trigger_keyword="max")

# Then just speak:
"max, what's the system status?"
"max, backup the database" 
"max, check for security updates"

🧰 Built-in Tools (50+)

🎯 Essential Tools - Core functionality

Tool	Purpose	Example
shell	Execute commands with real-time output	check disk usage
editor	File editing with syntax highlighting	modify config files
python_repl	Interactive Python execution	run data analysis scripts
http_request	Universal HTTP client	call any REST API

🎙️ Voice & Audio - Speech interaction

Tool	Purpose	Example
listen	Speech-to-text with trigger keywords	"max, what time is it?"
speak	Text-to-speech (multiple engines)	convert text to speech
realistic_speak	Natural speech with emotions	"[S1] hello! (laughs)"

🧠 Memory & Data - Information systems

Tool	Purpose	Example
sqlite_memory	Local memory with SQL queries	search conversation history
memory	Cloud knowledge base (Bedrock)	semantic search across docs
sql_tool	Universal database client	connect to PostgreSQL/MySQL
data_viz_tool	Create charts and visualizations	generate sales reports

🌐 Integrations - External services

Tool	Purpose	Example
use_github	GitHub GraphQL API	create issues and PRs
slack	Team communication	send messages and alerts
use_aws	AWS service integration	manage EC2, S3, Lambda
scraper	Web scraping	extract data from websites

🚀 Advanced - Power user features

Tool	Purpose	Example
create_subagent	Distributed AI via GitHub Actions	delegate complex tasks
load_tool	Dynamic tool loading	install community tools
tasks	Background processes	run monitoring scripts
workflow	Complex automation	orchestrate deployments

⚙️ Configuration

Quick Settings

# Use different AI providers
MODEL_PROVIDER=anthropic maxs
MODEL_PROVIDER=openai maxs
MODEL_PROVIDER=ollama maxs

# Enable specific tools only  
STRANDS_TOOLS="listen,speak,sql_tool,github" maxs

# Enable all tools
STRANDS_TOOLS="ALL" maxs

Team Collaboration (Advanced)

# Enable team features
export STRANDS_TOOLS="event_bridge,sql_tool,memory"
export AWS_REGION=us-west-2
export MAXS_EVENT_TOPIC=my-team

# Shared knowledge base
export STRANDS_KNOWLEDGE_BASE_ID=team-kb-id

External Services

# GitHub integration
export GITHUB_TOKEN=your-token

# Slack integration  
export SLACK_BOT_TOKEN=xoxb-your-token
export SLACK_APP_TOKEN=xapp-your-token

# AWS services
export AWS_REGION=us-west-2

💡 Advanced Examples

🔥 Hot Reload Tool Creation

Create and use tools instantly:

# Save to ./tools/crypto.py
from strands import tool
import requests

@tool
def crypto(coin: str) -> str:
    """Get cryptocurrency price."""
    response = requests.get(f"https://api.coinbase.com/v2/exchange-rates?currency={coin}")
    price = response.json()["data"]["rates"]["USD"]
    return f"{coin}: ${price}"

Then immediately:

maxs "get crypto price for bitcoin"  # Tool works instantly!

📊 Data Analysis Pipeline

# Complete workflow in one command
maxs "connect to PostgreSQL, analyze user growth trends, create visualization, and save report to S3"

# Or step by step
maxs "connect to database 'users' on localhost"
maxs "create monthly signup chart for last 6 months" 
maxs "export chart as PNG and upload to S3 bucket"

🤖 Multi-Agent Workflows

# Deploy specialized agents
maxs "create subagent for security audit with GitHub Actions"
maxs "create subagent for performance testing with detailed metrics"

# P2P agent communication (no internet needed)
maxs "start bitchat and enable agent triggers"
# Agents automatically coordinate and share results

🛡️ Privacy & Security

Local-first by design:

• Core functionality works completely offline
• Conversation history stored locally in /tmp/.maxs/
• Custom tools saved in ./tools/ directory
• No external data transmission except to chosen AI provider

Optional cloud features (when enabled):

• AWS Bedrock for knowledge base (requires AWS credentials)
• Team collaboration via EventBridge (optional)
• External APIs only when explicitly used (GitHub, Slack, etc.)

🔧 Troubleshooting

Common Issues

AI Provider Problems:

# Try local model as fallback
ollama serve && ollama pull qwen3:4b
MODEL_PROVIDER=ollama maxs

Voice Recognition Issues:

maxs
> listen(action="list_devices")  # Check available microphones

Tool Loading Problems:

# Reset and enable all tools
STRANDS_TOOLS="ALL" maxs

Database Connection Issues:

maxs
> sql_tool(action="connect", database_type="postgresql", host="localhost")

🚀 Why Developers Choose maxs

"Started using maxs for quick scripts. Now our entire team builds tools by talking to them. Saved 20 hours last week alone."
— DevOps Engineer

Key advantages:

• ⚡ Instant gratification - Working tools in 30 seconds
• 🎙️ Voice-first - Build while you think out loud
• 🔄 Hot reload - Iterate without interruption
• 📡 Distributed - Team coordination without servers
• 🧠 Intelligent - Remembers your context and decisions

---

📦 Installation

pipx install maxs && maxs

Alternative setups:

• Development: git clone https://github.com/cagataycali/maxs && pip install -e .
• Binary: pip install maxs[binary] then pyinstaller --onefile -m maxs.main

📄 License

MIT - Use it however you want!