NanoClaw - AI Agent on Raspberry Pi
NanoClaw is an open-source, secure AI agent framework (~3,900 lines across 15 files) that I deployed on a Raspberry Pi 3 collecting dust in a drawer. It runs as a Discord bot with a distinct personality, scanning grocery receipt photos, extracting every line item (merchant, quantity, price per kg, total), and storing everything locally in SQLite. No cloud. No subscription. Send a receipt photo on Discord and she replies with witty commentary like 'Two kinds of onions and two bags of rice - you were PREPARING for something' while quietly building a price-per-kg database across every shop. It tracks most-bought items, spending patterns, hall-of-fame moments, and acts as a personal inflation tracker built from actual receipts, not government averages. Costs ~$1/month to run. Your data never leaves your desk.
NanoClaw - AI Agent on Raspberry Pi
Frontend
Backend
Database
Tools
- • Running a capable AI agent on a Raspberry Pi 3 with only 1GB RAM
- • Accurately extracting line items from grocery receipt photos via OCR
- • Giving the agent a genuine personality beyond generic assistant responses
- • Deployed and customized NanoClaw on a Raspberry Pi 3 for grocery receipt scanning
- • Configured the agent with a distinct personality and witty contextual responses
- • Built a price-per-kg tracking system stored locally in SQLite
Runs 24/7 on a $35 Raspberry Pi 3 for ~$1/month with all data stored locally
Built a personal inflation tracker from real receipts across multiple shops
Tracks spending patterns, most-bought items, and price changes over time
Other Projects
Chain LLM steps into a workflow with {{stepId.output}} substitution between steps. Pick a preset, edit any prompt, watch the chain execute step by step.
View ProjectCLI + web viewer that runs your prompts against test cases with contains / regex / equals / llm-judge assertions, produces a JSON report, and fails the build on regressions in CI.
View Project