Step-by-step guides for running PicoClaw as a lightweight AI assistant on Raspberry Pi (including Pi 5) and Linux, plus Windows WSL2, Telegram and Discord bots, local LLMs (Ollama), Docker and Docker Compose, VPS and systemd, nginx HTTPS, Cloudflare Tunnel, Tailscale, GitHub Actions, n8n webhooks, and major LLM APIs (OpenAI GPT, Anthropic Claude, Google Gemini, Groq, DeepSeek, OpenRouter, OpenAI-compatible gateways). For long-form essays on edge AI, ops, and LLM strategy, see the PicoClaw blog.
Turn a \$10–\$35 Raspberry Pi into a persistent AI assistant that listens for events, calls cloud or local LLMs, and sends notifications.
Run PicoClaw as a self‑hosted AI agent on a VPS, NUC, or homelab server with cron‑driven jobs, webhooks, and local‑only data flow.
Wire Home Assistant automations into PicoClaw so a lightweight AI assistant can react to smart‑home events and send intelligent responses.
Run PicoClaw in Docker on your homelab server or NAS to centralise AI automations next to your other self‑hosted services.
Create AI‑powered cron jobs that call PicoClaw on a schedule to turn raw logs and metrics into readable summaries and alerts.
Run PicoClaw on Windows using WSL2. Install in Linux, configure providers, and run your lightweight AI assistant from the terminal.
Set up a Telegram AI assistant with PicoClaw using picoclaw gateway and your LLM provider keys.
Run a Discord AI bot with PicoClaw. Connect Discord messages to lightweight AI responses through the gateway.
Use PicoClaw with on-device or local models by connecting it to Ollama. Keep prompts and logs in your environment.
Tune PicoClaw on Raspberry Pi for smoother long-running assistant behavior. Reduce CPU spikes and optimize workloads.
Deploy PicoClaw on RISC-V hardware. Choose the correct release, configure providers, and run your lightweight AI assistant.
Pick the right LLM provider for a lightweight AI assistant. Compare OpenRouter and OpenAI for latency, cost, and model variety.
Deploy PicoClaw as an always-on assistant using Docker Compose with config volumes and reliable restarts.
Use Groq as your PicoClaw provider for low-latency assistants, bots, and cron jobs—especially on small VPS or Raspberry Pi front-ends.
Route PicoClaw to DeepSeek when you want strong models with aggressive pricing for high-volume summaries and alerts.
Point PicoClaw at LM Studio, vLLM, LiteLLM, or any OpenAI-shaped HTTP API for self-hosted models and internal gateways.
Trigger PicoClaw from n8n workflows for self-hosted AI steps next to hundreds of SaaS connectors.
Run PicoClaw under systemd for reboot-safe, production-style assistants on servers and Raspberry Pi OS.
Deploy PicoClaw on a cheap headless VPS: SSH, systemd, webhooks, and cron without a heavy Node or Python stack.
Use Gemini from AI Studio as your PicoClaw provider for multimodal models while keeping the agent lightweight on-device.
gemini provider slotRun Claude-backed assistants through PicoClaw on Pi, VPS, or Docker—quality from Anthropic, tiny footprint locally.
anthropic in providersDirect OpenAI integration for GPT models when you want the official API and billing without a heavy local stack.
Terminate TLS with Let’s Encrypt and proxy to PicoClaw for safe webhooks and public gateways.
Expose PicoClaw through cloudflared when you cannot open inbound ports—CGNAT-friendly homelabs.
Reach PicoClaw only on your tailnet: stable DNS between Pi, NAS, laptop, and GPU box.
Extra CPU and I/O headroom for gateway + automation while PicoClaw still uses minimal RAM.
Trigger LLM workflows from CI: push, schedule, or manual—thin runners posting to your PicoClaw API.
PicoClaw is designed for lightweight AI automation: scheduled tasks, webhook‑driven agents, and background assistants that run where typical chatbots cannot. Use these guides as starting points and adapt them for:
When you are ready to go deeper, visit the Docs and Configuration reference.