Learn it. Build it. Ship it for others.
Ollama routing tutorial for Claude Code
This repo is a visual guide for splitting work between Claude Desktop on Anthropic and Claude Code routed through Ollama. The goal is to keep planning and review in Claude while pushing repetitive terminal work to a free open-source model. The deck is self-contained HTML, and the setup prompt is written so Claude can do most of the configuration for you.
Builders who use Claude Code and want a cheaper setup for repetitive terminal work.
You can keep Claude Desktop for strategy while moving heavy Claude Code tasks to Ollama and reducing quota usage.
What it does
Two-engine setup
Shows Claude Desktop handling higher-level work while Claude Code runs through Ollama in the terminal.
Copy-paste setup prompt
Provides a prompt that automatically detects the OS, installs the pieces, configures the router, and checks the result.
Side-by-side verification
Ends with a check that both engines are working so you can confirm the setup succeeded.
Self-contained slide deck
Ships as plain HTML that opens in any browser with no build step.
OS-aware walkthrough
Covers macOS, Windows with WSL2, and Linux.
How to get it
- 1Run
git clone https://github.com/Coherence-Daddy/use-ollama-to-enhance-claude.git cd use-ollama-to-enhance-claude/presentation open index.html # macOS # or just drag index.html into a browser
README
Use Ollama to Enhance Claude — Two-Engine Setup
Pair Claude Desktop on Anthropic with Claude Code routed through Ollama in your terminal. Strategy stays on Pro. Heavy footwork runs on a free open-source model. Cut your Claude Code bill ~90%.
What this is
A 21-slide visual walkthrough that pairs your Anthropic Claude Desktop app with an Ollama-backed Claude Code terminal session — so the strategic work stays on Pro while the heavy lifting runs on a free open-source model (Gemma, Qwen, DeepSeek, your pick).
- Auto-detects your OS — macOS, Windows + WSL2, Linux
- Copy-paste prompt — drop one block into Claude and it does ~98% of the setup for you
- Verifies both engines side-by-side at the end so you know it actually worked
- Self-contained HTML — no build step, opens in any browser
Why this exists
Claude Pro on the Desktop app is great for thinking, planning, and architecture. Claude Code in the terminal eats quota fast — context-heavy tasks like lints, refactors, file batch ops, and grep-and-replace can burn through your monthly limit in days.
The fix: route Claude Code through Ollama (local or cloud-hosted free model). Two engines, same UX, one bill cut by an order of magnitude.
Quick start
1) Open the live tutorial (recommended)
The hosted version has the full visual deck, OS-aware steps, and the copy-paste prompt:
→ coherencedaddy.com/tutorials/use-ollama-to-enhance-claude
2) Or use the copy-paste prompt directly
If you want to skip the visuals and let Claude do it all for you:
- Open
prompts/copy-paste-prompt.md - Copy the entire file contents
- Paste into a fresh Claude Desktop / Claude.ai conversation
- Follow along — it auto-detects your OS, installs everything, configures the router, and verifies both engines
3) Or run the presentation locally
git clone https://github.com/Coherence-Daddy/use-ollama-to-enhance-claude.git
cd use-ollama-to-enhance-claude/presentation
open index.html # macOS
# or just drag index.html into a browser
What you'll have when you're done
| Engine | Where it runs | What it's for |
|---|---|---|
| Claude Desktop (Anthropic) | Native app | Strategy, architecture, code review, tricky bugs |
| Claude Code → Ollama | Your terminal | Lints, refactors, repetitive edits, file batch ops |
Two side-by-side panes. Same UX. One of them is free.
What's in this repo
.
├── README.md ← this file
├── LICENSE ← MIT
├── prompts/
│ └── copy-paste-prompt.md ← the canonical setup prompt
└── presentation/
├── index.html ← 21-slide visual walkthrough
├── cd-face-coral.png ← brand asset
└── copy-paste-prompt.md ← (mirror of /prompts version, kept beside the deck)
The presentation is the same exact HTML served at coherencedaddy.com/tutorials/use-ollama-to-enhance-claude. Open it locally, host it yourself, or fork it for your own walkthroughs.
Companion tutorial
If you want Claude Desktop wired to your Obsidian vault as a permanent semantic memory (with a one-click installer), see the sibling repo:
→ Coherence-Daddy/give-obsidian-a-memory — Local brain over your notes, indexed by BGE-M3, queryable by meaning. Ships with a one-click installer.
The two tutorials are independent — pick either or run both. This one is the lighter "two-engine" terminal setup; the other is the full Obsidian + local brain stack.
Topics
claude-code ollama gemma llm-tools cost-optimization agentic-coding anthropic open-source-llm
License
MIT — see LICENSE. Use it, fork it, ship a YouTube walkthrough of it. No attribution required, but if you do credit it, link to coherencedaddy.com.
About Coherence Daddy
Coherence Daddy is a 508(c)(1)(A) faith-driven technology organization on a mission to help humanity be more coherent. We build private, secure self-help tools — and the occasional tutorial that saves you a few hundred bucks on AI bills.
- Website: coherencedaddy.com
- Free tools (523+): coherencedaddy.com/tools
- Tutorials: coherencedaddy.com/tutorials
- X / Twitter: @coherencedaddy
If this saved you a real chunk of money, the kindest thing you can do is ⭐ this repo and share the tutorial link.
Files in the repo
- presentation
- prompts
- .gitignore
- LICENSE
- README.md
Discussion (0)
Ask about usage, or say what you built with itSign in to join the discussion.
No comments yet. Be the first to say what this is good for.
More tutorials & guides
Bash is all you need - A nano claude code–like 「agent harness」, built from 0 to 1
A visual, example-driven guide to Claude Code — from basic concepts to advanced agents, with copy-paste templates that bring immediate value.
Specification and documentation for Agent Skills

💻 vibe coding 101|The first course for AI-native product builders.

End-to-end, code-first tutorials for building production-grade GenAI agents. From prototype to enterprise deployment.