- Quick lookups and simple edits run on a fast, low-cost model
- Long research and hard reasoning go to your strongest model
- Describe each route in plain words and Hermes picks the right one per task
- Every result shows the route, provider and model that did the work
Advanced Delegation setup
Transcript
Music, with a narrator reading each step; the captions show the same words. Spoken lines are quoted with who says them; the rest describes what's on screen.
- At 0:00
Title card: "Give every Hermes subagent its own model." Beside it, a Hermes Desktop window shows its subagent settings.
- At 0:04
"Here's how."
- At 0:05
Hermes Desktop on a new session, with the "HERMES AGENT" wordmark.
- At 0:06
Step 1 of 5, "Open Settings". The keys Ctrl + , appear, and a narrator says: "Open Settings. Control comma works from anywhere."
- At 0:07
Settings opens on Model, Main model: provider "ChatGPT or Codex Subscription", model gpt-6.1-sol, reasoning High.
- At 0:10
Step 2, "Model, then Subagents". The narrator: "Go to Model, then Subagents." A callout points to Subagents.
- At 0:11
The Subagents page, under Routing: Default (in use), "Every subagent runs on the subagent model below"; Model routes, "Name a few models. The agent picks one for each subagent"; and Profile routes, "Subagents run as one of your profiles, with its own model, skills and keys". Both route modes are tagged Plugin, and a line says they come from the Advanced Delegation plugin, next to an "Install plugin" button.
- At 0:12
Step 3, "Turn on Model routes". The narrator: "Choose Model routes. The first time, Hermes installs the Advanced Delegation plugin for you."
- At 0:16
A callout, "One click", points to "Install plugin".
- At 0:17
A dialog: "advanced-delegation", tagged Official, "Reviewed by Nous. Installs for default and turns on right away.", with Cancel and Install buttons.
- At 0:18
The narrator: "One click." The cursor clicks Install.
- At 0:21
Back on Subagents, the cursor clicks Model routes.
- At 0:22
Model routes is now in use: "No routes yet. Until you add one, subagents use the main agent's model."
- At 0:23
Step 4, "Add a fast route". The narrator: "Add a route. Call it fast, say when to use it, and pick a quick, low-cost model." The cursor clicks Add route.
- At 0:26
The New model route form: "The agent reads the name and description to decide which route a subagent takes." It has Name, When to use it, Provider, Model, and optional Reasoning effort and Speed.
- At 0:28
Name: "fast".
- At 0:30
When to use it: "quick lookups and simple edits". Provider: ChatGPT or Codex Subscription.
- At 0:33
The Model list offers gpt-6.1-sol and Custom model, and Custom model is picked.
- At 0:39
"gpt-6-luna" is typed in as the model.
- At 0:40
Add route is clicked, and fast is listed as the default route: "quick lookups and simple edits", ChatGPT or Codex Subscription, gpt-6-luna.
- At 0:42
Step 5, "Add a deep route". The narrator: "Add a second route, call it deep, with your strongest model and high reasoning."
- At 0:45
Name: "deep".
- At 0:48
When to use it: "hard reasoning and long research". Provider: ChatGPT or Codex Subscription.
- At 0:51
Model: gpt-6.1-sol.
- At 0:54
The Reasoning effort list opens: Subagent default, none, minimal, low, medium, high, xhigh, max and ultra. High is picked.
- At 0:59
Add route is clicked.
- At 1:00
Two routes are listed: fast (Default) on gpt-6-luna, and deep, "hard reasoning and long research", on gpt-6.1-sol with reasoning effort high. A note says subagents that pick no route use fast.
- At 1:02
Caption: "Two routes: fast and deep".
- At 1:07
The narrator sums up: "Settings, Model, Subagents, Model routes."
- At 1:08
Recap: 1 Open Settings, 2 Model, then Subagents, 3 Turn on Model routes, 4 Add a fast route, 5 Add a deep route.
- At 1:15
End card: "@HermesWatcher".
Advanced Delegation lets the agent pick a named model route for each subagent from routes you define, and only the route names and their one-line descriptions reach the model. In the video it is set up in Hermes Desktop under Settings, Model, Subagents: choosing Model routes installs the plugin, then a fast route gets a low-cost model and a deep route gets the strongest model with high reasoning.
Needs Hermes 0.22.5 or newer. In the terminal, hermes delegation config sets up the same routes.
Install from the catalog
hermes plugins install advanced-delegationIt asks whether to turn the plugin on: say yes, then restart Hermes if it tells you to. The bare name installs the reviewed, pinned catalog version.
Plugin page Watch on X (opens in a new tab)More plugin videos
The post on X
The post as written on Nov 4, 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
Hermes Android
Pair an Android phone and Hermes can read its screen, tap, text and call, with your OK before anything sends. Transcript
Per the catalog listing, Hermes Android gives Hermes 58 android_ tools for a paired Android phone through the FlipsiBridge app: reading the screen, tapping and typing, the camera, screenshots, files, notifications, texts and calls. The phone dials out to a relay you host, so it works behind NAT with no port forwarding or VPN. In the video, Hermes finds a contact and sends a text once it's allowed, turns on the flashlight, reads the screen and starts a timer. Transcript
TimelinesAI WhatsApp
Hermes finds who's waiting in a team WhatsApp inbox, labels and assigns chats, and sends replies you approve. Transcript
Per the catalog listing, this plugin connects Hermes to TimelinesAI's hosted WhatsApp server with a browser sign-in: it searches and summarises chats, finds unanswered ones, and sends messages, notes and labels, and its bundled skill makes Hermes confirm the recipient and exact text before any send. In the video, Hermes finds three chats waiting two days or more, labels and assigns them, and sends a follow-up once it gets an OK. Transcript
Imagination Studio
Hermes makes pictures, clips, voices, songs and sound effects through a studio running on your own machine. Transcript
Per the catalog listing, Imagination Studio connects Hermes to the Imagination Studio app on your own machine, which renders pictures, video clips with sound, voice lines, songs and sound effects and sends the files back to the chat. In the video the plugin is found and installed under Plugins in Hermes Desktop, /studio opens it, and the Studio's story, shot planning, looks, edit and sound pages appear between cinematic action clips. Transcript