May 23, 2026
Claw and Order
Show HN: I built a RAG and knowledge graph agent that runs locally
Privacy-first laptop AI drops, but commenters immediately ask: where’s the code
TLDR: Claw-Coder promises a private AI coding assistant that runs on your laptop and tests its own work safely. Commenters liked the privacy pitch but immediately challenged the novelty, questioned the closed-source rollout, and turned the thread into a classic “show us the receipts” showdown.
A new Hacker News launch is pitching a very juicy promise: an AI coding helper that stays entirely on your laptop, so your private code doesn’t get shipped off to big cloud companies. That alone was enough to grab attention. The creator says the tool, called Claw-Coder, boosts weaker local AI models by giving them extra help: a kind of smart map of your code, a searchable memory system, web search, and a safe testing area so it can try code without trashing your machine.
But the real fireworks were in the comments, where the community instantly split into “cool idea” and “absolutely not buying this yet.” One commenter basically said: hold on, the big-name coding tools already support local models, so this may not be as revolutionary as advertised. Another threw a sharper jab: if the project is closed source and the first thing people see is a Homebrew install command, that feels... weird. In internet terms, that’s the kind of side-eye that can start a whole thread.
Then came the comedy. One commenter zoomed in on the phrase “ever had of RAG” like a grammar sniper, while another cut through all the hype with a brutally practical question: will this save tokens, or not? Even the creator’s own comment — asking whether a “knowledge graph” makes AI reason better about code — read like tossing raw meat into a debate pit. The vibe? People love the privacy angle, but they want receipts, not just promises.
Key Points
- •Claw-Coder is presented as a local AI coding agent intended to run on a user's laptop rather than relying on cloud-hosted coding agents.
- •The article says the tool was created to address privacy and security concerns by keeping code, RAG data, and knowledge graph information local.
- •The author argues that local language models need added tools to offset performance limits compared with larger cloud models.
- •Claw-Coder is described as using a knowledge graph and RAG with a vector store to help local models understand large codebases and code relationships.
- •The post says the agent includes a search tool, Docker-based code execution in an isolated workspace, and browser-output inspection for HTML and CSS via a vision LLM.