Advanced Delegation

Plugin · Official, by Nous Research ·  · 0:26

Transcript

Music, no speech. Track: Infraction, "Ready, Steady, Go". Below: what's on screen, beat by beat.

  1. At 0:00

    Title: "Every subagent gets its own model." Two route cards appear: "fast" pointing to "gpt-6-luna" and "deep" pointing to "gpt-6.1-sol", then a table pairs each route with its actual model.

  2. At 0:04

    "Set it up once." labelled "Step 1 · In the terminal". Windows PowerShell types "hermes delegation".

  3. At 0:06

    Yellow screen: "Pick Model routes." The menu offers Default, Model routes and Profile routes. Choice 2 is made: "Model routes: the agent picks one per subagent".

  4. At 0:08

    "Name a route": "fast". The prompt asks for a route name and a description, "one line the agent reads".

  5. At 0:09

    "Pick its model." A provider list shows NVIDIA NIM, Nous Portal, OpenRouter, OpenCode Go, Google AI Studio and a custom endpoint, and option 1 is chosen.

  6. At 0:10

    "fast" is set to "gpt-6-luna". The terminal shows "Saved route 'fast'." with the description "cheap and quick: lookups, summaries, simple edits."

  7. At 0:12

    Blue screen: "Add another": "deep". It gets "gpt-6.1-sol" from a short model list and the description "slow and careful: hard debugging, research and design".

  8. At 0:15

    "Routes are live." fast is picked as the default route, and the terminal confirms "Delegation mode: Model routes (fast (default), deep)."

  9. At 0:17

    Green screen: "Then just ask." labelled "Step 2 · Hermes Desktop or CLI". A message asks Hermes to use fast for the last 5 Hermes releases and deep for the one change that matters most to plugin devs.

  10. At 0:20

    "fast is done. deep keeps thinking." Hermes reports "Fast finished on gpt-6-luna" while the deep route is still reviewing the most consequential plugin change.

  11. At 0:22

    "Every result shows its route." A table lists fast on gpt-6-luna and deep on gpt-6.1-sol, both with the provider openai-codex.

  12. At 0:24

    The video closes on "@HermesWatcher" on a cream screen.

Lets each subagent run on a model route you define, like fast, deep or local, or as one of your other Hermes profiles. Only route names and descriptions reach the model.

Needs Hermes 0.22.5 or newer.

The post on X

HermesWatcher @HermesWatcher ·
Okay, this one is sneaky good.Hermes can now run every subagent on a different model.Advanced Delegation is an official Nous Research plugin. You set up named model routes, like fast and deep, and Hermes sends each piece of work down the route that fits.
  • A fast, cheap model for lookups and summaries
  • Your strongest model for the hard thinking
  • A local model on your own endpoint
  • Or another Hermes profile, with its own model, skills and SOUL.md
Every result shows the route and model it actually ran on.And Hermes only sees the names and descriptions you write, so it can't invent a route.One agent, a whole bench of models.

The post as written on Oct 23, 2026, with any sign-up offers and pointers to replies left out. Figures in it are from that day. The summary above follows the official docs, so check the docs (opens in a new tab) for the version you run.

More videos

Office Desk

Plugin · Community, by DECRUX9812 · 

Edits Word, Excel and PowerPoint files, with tracked changes in Word. Every write is backed up and undoable. Transcript

Hermes edits Word, Excel and PowerPoint files with tracked changes, recalculated formulas and in-place slide edits. Every write gets a backup and an undoable receipt you review in Hermes Desktop. Transcript

All videos