Your ultra-lightweight bot deserves professional cloud infrastructure. Deploy to managed hosting with zero maintenance overhead.
Comprehensive documentation for configuring picoclaw across platforms, skills, and advanced features
Pico Claw uses a simple YAML configuration file to control all aspects of the AI assistant. This guide covers everything from basic setup to advanced tuning.
~/.picoclaw/config.yamlpicoclaw config validate# config.yaml - Pico Claw Configuration
api:
provider: anthropic
key: ${ANTHROPIC_API_KEY}
model: claude-3-5-sonnet-20241022
platforms:
telegram:
enabled: true
token: ${TELEGRAM_BOT_TOKEN}
memory:
persistent: true
max_size: 2MB
skills:
enabled: true
auto_install: false
logging:
level: info
file: /var/log/picoclaw.log
Essential configuration settings to get picoclaw up and running. API key setup is required before first use.
# config.yaml - API Configuration
api:
provider: anthropic # Anthropic Claude
key: ${ANTHROPIC_API_KEY} # Use environment variable
model: claude-3-5-sonnet-20241022
max_tokens: 4096
temperature: 1.0
# Set API key via environment
export ANTHROPIC_API_KEY="sk-ant-api03-..."
# Or in config.yaml (not recommended)
api:
key: "sk-ant-api03-your-actual-key"
Security: Always use environment variables for API keys, never commit keys to version control
# Recommended models for Pico Claw
api:
# High performance (default)
model: claude-3-5-sonnet-20241022
# Cost-effective option
# model: claude-3-haiku-20240307
# Maximum capability
# model: claude-3-opus-20240229
# General Pico Claw settings
general:
name: "Pico Claw"
language: en
timezone: UTC
# Response behavior
response:
max_length: 2000 # Character limit
typing_indicator: true
retry_on_error: 3
# System behavior
system:
auto_update: false
telemetry: false # Respect privacy
debug_mode: false
Pico Claw supports multiple messaging platforms simultaneously. Configure Telegram, Discord, QQ, DingTalk, Feishu, and WhatsApp integrations.
# config.yaml - Telegram Setup
platforms:
telegram:
enabled: true
token: ${TELEGRAM_BOT_TOKEN}
# Optional: Advanced settings
allowed_users: [] # Empty = all users
admin_users: [123456] # User IDs with admin access
max_message_length: 4096
# Commands
commands:
start: "Hello! I'm Pico Claw AI assistant."
help: "Available commands: /start, /help, /status"
# Features
features:
inline_mode: true
group_chat: true
file_upload: true
Get token from @BotFather on Telegram
# config.yaml - Discord Setup
platforms:
discord:
enabled: true
token: ${DISCORD_BOT_TOKEN}
# Bot settings
prefix: "!" # Command prefix
status: "online"
activity: "Helping users"
# Permissions
allowed_guilds: [] # Empty = all servers
allowed_channels: [] # Empty = all channels
# Features
features:
slash_commands: true
message_commands: true
mentions: true
dm_support: true
Create bot at Discord Developer Portal
# config.yaml - QQ Setup
platforms:
qq:
enabled: true
protocol: "shamrock" # or "go-cqhttp"
# Connection
host: "127.0.0.1"
port: 8080
access_token: ${QQ_ACCESS_TOKEN}
# Bot account
qq_number: 123456789
# Features
features:
group_chat: true
private_chat: true
admin_mode: false
Requires QQ bot framework (Shamrock or go-cqhttp)
# config.yaml - Enterprise Platforms
platforms:
dingtalk:
enabled: true
app_key: ${DINGTALK_APP_KEY}
app_secret: ${DINGTALK_APP_SECRET}
feishu:
enabled: true
app_id: ${FEISHU_APP_ID}
app_secret: ${FEISHU_APP_SECRET}
whatsapp:
enabled: false # Configuration only
# WhatsApp setup requires additional infrastructure
Enterprise platforms require app registration
Run Pico Claw on multiple platforms simultaneously - the same instance can serve Telegram, Discord, and QQ users concurrently.
Each platform adds <2MB RAM - picoclaw can run 3+ platforms on devices with 64MB total memory.
Single config.yaml controls all platforms - change settings once, apply everywhere.
Fine-tune picoclaw performance with memory management, persistent storage, custom prompts, and logging controls.
# Memory optimization for Pico Claw
memory:
# Persistent memory
persistent: true
storage_path: ~/.picoclaw/memory
max_size: 2MB
# Runtime memory
max_conversations: 10
max_messages_per_conversation: 100
# Cleanup
auto_cleanup: true
cleanup_interval: 24h
retention_days: 7
# Customize Pico Claw behavior
prompts:
system: |
You are Pico Claw, a helpful AI assistant
running on a RISC-V device with limited resources.
Be concise and efficient.
personality: friendly
response_template: |
{{ content }}
---
Powered by Pico Claw
# Pico Claw logging settings
logging:
level: info # debug, info, warn, error
# File logging
file:
enabled: true
path: /var/log/picoclaw.log
max_size: 10MB
max_backups: 3
compress: true
# Console logging
console:
enabled: true
format: json # json or text
# Optimize Pico Claw performance
performance:
# Concurrency
max_concurrent_requests: 5
request_timeout: 30s
# Caching
cache:
enabled: true
ttl: 5m
max_entries: 100
# Rate limiting
rate_limit:
requests_per_minute: 20
burst: 5
Extend picoclaw functionality with skills installed from GitHub repositories. Manage skill permissions and create custom skills.
# config.yaml - Skills Configuration
skills:
enabled: true
auto_install: false
skills_directory: ~/.picoclaw/skills
# Installed skills
installed:
- name: weather
repo: github.com/picoclaw/skill-weather
version: v1.0.0
enabled: true
- name: calculator
repo: github.com/picoclaw/skill-calculator
version: v1.2.0
enabled: true
- name: image-gen
repo: github.com/picoclaw/skill-image
version: v0.5.0
enabled: false # Disabled but installed
Install skills with: picoclaw skill install github.com/picoclaw/skill-name
# Manage Pico Claw skills
# Enable skill
picoclaw skill enable weather
# Disable skill
picoclaw skill disable image-gen
# List all skills
picoclaw skill list
# Update skill
picoclaw skill update weather
# Remove skill
picoclaw skill remove calculator
# Per-skill permissions in config.yaml
skills:
permissions:
weather:
network: true # Allow internet access
files: false # No file system access
users: [] # All users (empty = all)
calculator:
network: false
files: false
users: []
admin-tools:
network: true
files: true
users: [123456] # Admin users only
Create custom skills for Pico Claw:
~/.picoclaw/skills/custom/Protect your picoclaw deployment with API key security, user permissions, data privacy settings, and network controls.
# Secure API key management
# Use environment variables
export ANTHROPIC_API_KEY="sk-ant-..."
export TELEGRAM_BOT_TOKEN="123456:ABC-..."
# In config.yaml
api:
key: ${ANTHROPIC_API_KEY}
# File permissions
chmod 600 ~/.picoclaw/config.yaml
# Key rotation
picoclaw config update-key --api
Never commit API keys to version control
# Access control for Pico Claw
security:
# User authentication
require_auth: true
# Whitelist
allowed_users:
- 123456789 # Telegram user ID
- 987654321
# Admin users
admin_users:
- 123456789
# Blacklist
blocked_users:
- 111111111
User IDs are platform-specific
# Privacy controls for Pico Claw
privacy:
# Data collection
store_messages: false
store_user_data: false
# Telemetry
telemetry_enabled: false
error_reporting: false
# Encryption
encrypt_memory: true
encryption_key: ${ENCRYPTION_KEY}
Pico Claw respects user privacy by default
# Network security for Pico Claw
security:
# HTTPS enforcement
require_https: true
# Webhook validation
validate_webhooks: true
webhook_secret: ${WEBHOOK_SECRET}
# IP filtering
allowed_ips: [] # Empty = all IPs
# Firewall integration
firewall:
enabled: true
deny_by_default: false
Configure firewall rules on your system
Maximize picoclaw efficiency with memory optimization, response time tuning, resource limits, and monitoring.
Minimize Pico Claw memory footprint for ultra-low-resource devices:
# Optimize for minimal memory
memory:
persistent: false # Disable persistent memory
max_size: 1MB # Reduce max memory
# Limit conversations
max_conversations: 5
max_messages_per_conversation: 50
# Aggressive cleanup
auto_cleanup: true
cleanup_interval: 1h
Achieves <8MB RAM usage on minimal config
Configure Pico Claw for optimal response speed:
# Optimize response time
performance:
# Reduce timeout
request_timeout: 15s
# Increase concurrency
max_concurrent_requests: 10
# Enable caching
cache:
enabled: true
ttl: 10m
preload_common: true
Balance speed with API costs and rate limits
Prevent Pico Claw from exceeding system resources:
# Resource constraints
limits:
# Memory limits
max_memory: 20MB # Hard limit
memory_warning: 15MB # Warning threshold
# CPU limits
max_cpu_percent: 50
# Disk limits
max_disk_usage: 100MB
# Connection limits
max_connections: 100
Picoclaw respects system resource limits
Track Pico Claw performance metrics:
# Performance monitoring
monitoring:
enabled: true
# Metrics collection
metrics:
memory_usage: true
response_time: true
api_calls: true
error_rate: true
# Export format
export:
prometheus: true
json: true
# Alerts
alerts:
memory_threshold: 18MB
error_rate_threshold: 5%
Integrate with monitoring tools like Prometheus
Explore complete configuration examples for different use cases: minimal devices, multi-platform deployments, and enterprise setups.