Who it is forA developer who wants cheap models on quick tasks and strong ones on hard reasoning.
The problemUsing one model for everything wastes money on easy steps and underpowers hard ones.
The outcomePer category model chains that fall back to the next model when a provider rejects one.
What you need first
- Hermes Agent installed
- macOS or Linux (install script) or Windows PowerShell 5.1+
- Accounts or keys for the model providers you want in each lane
Steps
- Run
omh modelto open the picker - Review the categories such as ultrabrain, deep, visual-engineering and quick
- Edit each chain and its reasoning effort
- Run a /dag-ulw goal and check which model ran each lane
Prompt for your agent
Open omh model in rlaope/oh-my-hermes and set <cheap model> for quick lanes and <strong model> for ultrabrain and deep lanes, then run a small /dag-ulw goal to confirm the routing.
Caveat. This card is written from the rlaope/oh-my-hermes README, not from a test run by Deduped. Before you start, make sure you have Hermes Agent installed. Check the repo for newer instructions before you rely on these steps.
Use this in your AI agent
Ask your agent
Using the Deduped MCP, get the setup steps for rlaope/oh-my-hermes and "Choose which model handles each kind of work"
Add Deduped to Claude Code once
claude mcp add --transport http deduped https://dedupedai.com/mcp
Every Monday: the 5 hottest repos and exactly how to use them. Free, one email a week, unsubscribe anytime.
Get them freeShare: X LinkedIn Hacker News Email Copy link
Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer