[Update 2026-02-02: nemotron-3-nano also performs well on same setup; see comment below]

This is a guide to setting up Clawdbot/Moltbot with a local Ollama model that actually works -- meaning it has good tool use and decent speed. The main requirement is 48GB of VRAM. I have yet to find a model that fits on less than this and still works on Moltbot.

The setup involves creating a tool-tuned variant of qwen2.5:72b and modifying a range of configs in Moltbot. At the end you'll get a local Moltbot instance that can use tools (exec, read, write, web search), read skills, and perform agentic tasks without any cloud API dependencies. On my system I get ~16 t/s and have yet to come across a tool/skill that my bot can't use.

Claude Opus wrote the first draft of this Gist, then I (a human) checked and edited it.

Terminal-based agentic coding assistants like Claude Code, Codex, Opencode, and Aider are generally used to work with codebases, but nothing prevents us from using them on other types of content.

Here is an example of how to use Aider on a collection of Foam notes. The basic idea can be transposed to any combination of coding assistant and note taking system.

1. Install Aider

2. Create a CONVENTIONS.md document and place it in the root of your Foam repo. Fill it with information about the types of notes you have, how you like to structure them, what styling conventions you follow, etc. Think of it as an extensive system prompt. If you're unsure what to include, describe your system in plain language to an LLM and get it to write a CONVENTIONS.md for you.