Guides & How-tos

OpenClaw's browsing skill turns your AI agent into a web scraper that can navigate pages, handle JavaScript rendering, and extract structured data without writing brittle CSS selectors. This guide walks through setting up the browsing tools, configuring proxy rotation for anti-bot evasion, building extraction workflows with memory and retry logic, and storing scraped results in a persistent workspace.

Heating and cooling consume 52% of residential energy in the US, but most smart thermostats follow rigid schedules that ignore occupancy and weather. An OpenClaw agent on a Raspberry Pi controls thermostats through Home Assistant with code-level safety bounds, making energy-aware decisions based on who is actually home. This walkthrough covers the hardware, ha-mcp skill setup, and occupancy patterns that reduce HVAC waste.

A geophone on a Raspberry Pi produces continuous seismic data, but most of what it records is foot traffic, trucks, and wind. OpenClaw adds a reasoning layer that distinguishes genuine earthquakes from local vibrations and sends alerts that explain what happened, where, and whether you should care.

rtl_433 decodes signals from over 320 wireless device protocols using a cheap USB radio dongle. Pairing it with an OpenClaw AI agent on a Raspberry Pi gives you anomaly detection, cross-sensor correlation, and automated responses on top of the raw telemetry. This guide covers the full stack from hardware to working agent.

PIR sensors miss people who sit still. mmWave radar does not. This guide covers wiring a 24GHz FMCW radar module to a Raspberry Pi, reading presence and distance data over UART or I2C, and running an OpenClaw agent that reasons about occupancy across zones, makes automation decisions, and logs events to a persistent workspace.

The AS3935 lightning sensor detects electromagnetic signatures from storms up to 40km away, but its raw interrupt output requires interpretation. By pairing it with OpenClaw on a Raspberry Pi, you can build an agent that tracks storm distance over time, predicts approach trajectories, and sends human-readable safety alerts through any messaging platform.

Fermentation temperature is the single biggest controllable variable in homebrewing. This guide covers wiring a DS18B20 probe to a Raspberry Pi, running an OpenClaw agent that reads temperature, reasons about fermentation stages, and adjusts heating or cooling relay outputs. It also covers logging brew data to a cloud workspace so you can review past batches and share results with your homebrew club.

Most Raspberry Pi EV charger projects stop at the protocol layer: read OCPP data, display it on a dashboard, done. This guide adds an AI reasoning layer. An OpenClaw agent running on the same Pi connects to your charger through Home Assistant's OCPP integration, monitors sessions in real time, and makes cost-optimized scheduling decisions based on electricity rates and solar generation.

AWS IoT Greengrass handles device provisioning, telemetry pipelines, and fleet management. OpenClaw adds AI reasoning on top. Running both on a Raspberry Pi gives you an edge node that collects sensor data, makes local decisions, reports to the cloud, and hands off results to your team through a shared workspace.

OpenClaw's built-in PDF tool lets agents extract text and analyze documents through native provider APIs or an automatic text-plus-image fallback. It supports up to 10 PDFs per call, page filtering for targeted extraction, and a 4 million pixel budget for image rendering. This guide covers both execution modes, batch processing workflows, structured extraction patterns, and how to persist extracted data in shared workspaces.

OpenClaw separates text and vision processing into independent model pipelines, letting you pair a fast text model with a specialized vision model in the same agent. You'll configure the imageModel setting, select the right vision model for your use case, and connect your agent to persistent storage for analysis results.

OpenClaw adds an intelligent automation layer on top of GitHub Actions. Instead of writing brittle if-then rules for every failure mode, you get an agent that monitors pipeline runs, classifies failures by type, reviews pull requests for real issues, and coordinates deployments across environments. This guide walks through setting up OpenClaw's CI/CD skills, connecting them to your GitHub repos, and storing pipeline artifacts in persistent workspaces.

OpenClaw agents can monitor inboxes, classify messages, extract data from attachments, and route actions through configurable skills. This guide walks through three integration paths (Gmail, AgentMail, Resend), covers scheduling and real-time triggers, and shows where to persist processed outputs so nothing gets lost between sessions.

OpenClaw turns a local AI agent into a customer support chatbot that monitors WhatsApp, Telegram, Slack, and dozens of other messaging channels. This guide walks through configuring OpenClaw's gateway, writing support-focused agent instructions, building a knowledge base for accurate answers, and routing complex issues to human agents. You keep full control of your data because everything runs on your own infrastructure.

OpenClaw stores agent configuration, credentials, conversation history, and workspace files in local directories that disappear if your disk fails or your container gets recycled. This guide walks through the built-in backup command, external encryption for production archives, automated scheduling, full restoration, and disaster recovery testing so you can recover a working agent in minutes instead of rebuilding from scratch.