I built a coding profile for Hermes Agent on Nous Portal, in two versions: portal-coder-premium and portal-coder-mid. a strong model plans, hands the edits to a low-cost subagent, then reviews the result and runs the tests itself. specs: - planner: Claude Sonnet 5.5 (premium) or GLM-5.3 (mid) - subagents: GPT-6 Luna - screenshots, compression and risky-command checks: GPT-6 Luna - titles and small side tasks: Ling 3.0 Flash - in a repo, Hermes switches to its coding toolset - approvals set to ask you before any force push - cost in the status bar it is for people who code with Hermes on the Portal and want a strong model making the calls, without paying its rate for every screenshot and title. every pick came from my own runs through the Portal API, with coding tasks judged by hidden tests. numbers are in each README. > premium hermes profile install > mid hermes profile install