Plan code changes with a strong model, edit with a cheap one
What it does
Hermes Agent runs coding jobs with a strong planner model paired with cheaper models for edits and side tasks. A top model (Claude Sonnet 5.5 or GLM-5.3) plans the changes, hands edits to a low-cost subagent (GPT-6 Luna), and the strong model reviews and runs tests.
This splits the cost so you get the strong model's judgment without paying its full rate for every screenshot or routine task.
Why it's worth it
Cuts costs by using an expensive model for critical planning decisions only, not routine work like screenshots and minor edits. For people working with Hermes on the Portal who want a strong model making decisions without paying its rate for every screenshot and title.

Step by step
Paste it into Hermes Agent and swap the highlighted parts for your own details.
Before you start Have your repo ready to work with and install portal-coder-premium or portal-coder-mid.
Get Hermes Agent to:
- Plan how to [implement the code changes I described].
- Present my plan for [how I'll make the changes] and wait for my approval.
- Execute the implementation using lower-cost models for edits and side work.
- Run the test suite and report the results.
- Tell me which models you used and any concerns you found.
Expected resultYour code changes are planned by a strong model, executed by a cost-effective subagent, reviewed by the strong model, and tested before you're notified.

