agentos.guide › blog
Hermes Advanced Delegation: The Quick Start, And Whether It Belongs In Your Agent OS
Hermes advanced delegation is an official Nous Research plugin that lets Hermes Agent send each subagent task to a named model route or to one of your Hermes profiles.
It is free, it plugs into the delegate_task tool, and it needs Hermes 0.22.5 or newer.
Without it, every subagent in a batch shares one model.
This page is the quick start, plus my honest answer on when it's worth adding to an Agent OS.
The Hermes advanced delegation quick start
This is the shortest path from built-in delegation to routed delegation inside your Agent OS.
| Step | What you do | Command or setting |
|---|---|---|
| 1 | Update to Hermes 0.22.5 or newer. | hermes update |
| 2 | Install the plugin in the profile that delegates. | hermes plugins install advanced-delegation |
| 3 | Pick Model routes or Profile routes and build the list. | hermes delegation config |
| 4 | Check what you saved. | hermes delegation config --show |
| 5 | Start a fresh session so the agent sees the routes. | /new, or /delegation <mode> --now |
The rest of this page explains what you get, whether it's worth it for your setup and how it fits each Agent OS profile.
What Hermes advanced delegation is
The official plugin page describes it as an official Hermes Agent plugin that adds two delegation modes to delegate_task, next to the built-in default.
delegation.backend: model_map, give each task a choice of named model routes that you define.
Profile routes, set with delegation.backend: profile_routes, let each task run as one of your Hermes profiles, with that profile's model, credentials, skills and SOUL.md.
Only one mode is active at a time, and each mode keeps its own settings when you switch.
Only the names and descriptions you write reach the model, so the agent can't see keys or invent a route.
Hermes core still runs the whole subagent lifecycle, and the plugin only decides which route each task takes.
Why it exists
The built-in default gives every subagent in a batch the same model.
The delegation docs say delegate_task has no per-task model parameter with the default strategy.
They also say a parallel batch of subagents typically burns the large majority of a run's tokens.
So a batch that mixes one hard task with nine easy ones either overpays or underpowers.
The v0.22.5 release on GitHub, published on 23 October 2026, added the pluggable delegation backends this plugin uses.
Step 1 to 4: install and build your routes
The plugin needs Hermes 0.22.5 or newer, and the plugin page says 0.22.0 to 0.22.4 refuses to install or enable it.
These are the install and setup commands from the plugin page.
hermes plugins install advanced-delegation hermes delegation config # pick the mode, then edit its routes hermes delegation config --show # print the current mode and routes (no terminal needed)
Desktop users can click Install plugin under Settings → Model → Subagents instead.
Plugins install per profile, so install it in the profile that delegates, while target profiles only need to exist with a working model.
The model routes editor uses hermes model's own pickers and validates each route before saving it.
The profile routes editor is a checklist of your other profiles, an optional description each and a default.
🔥 Want the exact Hermes Agent setup I run inside my Agent OS?
Inside the AI Profit Boardroom you get the installable Agent OS, step-by-step video tutorials, four coaching calls a week and 3,400+ members building real automations.
Model routes in YAML
If you'd rather edit the file, this is the shape from the plugin page, trimmed to two routes.
delegation:
backend: model_map
backends:
model_map:
options:
- name: fast # what the agent picks
description: "Cheap and quick: lookups, summaries, simple edits."
provider: openrouter # provider and/or base_url, plus model
model: google/gemini-3-flash-preview
- name: deep
description: "Slow and careful: hard debugging and design work."
provider: anthropic
model: claude-opus-5
reasoning_effort: high
default: fast # optional: route for tasks that pick none
Each route needs name, description and model, plus provider or base_url.
Optional keys are reasoning_effort, service_tier, fallback_providers, api_key and api_mode.
The page recommends provider: over api_key:, because a key written here is stored in plain text.
Profile routes in YAML
This is the profile routes shape from the same page.
delegation:
backend: profile_routes
backends:
profile_routes:
profiles: # allowlist AND what the agent reads about each profile
- name: coder
description: Writes and refactors code; runs tests.
- name: researcher
description: Web research and long-form summaries.
default: coder # optional: profile for tasks that pick none
The list is also an allowlist, so profiles you don't list can never be picked.
The plugin page warns that a listed profile's credentials, credential pool, SOUL.md and terminal policy become usable by the delegating profile's subagents.
Step 5: switch modes from any chat
/delegation # current mode, what each mode routes to, a picker to switch /delegation model_map # save it; it takes effect from the next session (/new) /delegation profiles --now # save it AND apply it to this chat now /delegation default # back to the built-in behaviour (from the next session)
Changes apply from the next session, because the tool list is part of the prompt cache.
--now rebuilds the current chat's tools at your next message, at the cost of one prompt-cache break.
On Telegram, Discord and Slack, bare /delegation shows buttons, while the CLI, TUI and Desktop use their own picker.
🔥 Want the exact Hermes Agent setup I run inside my Agent OS?
Inside the AI Profit Boardroom you get the installable Agent OS, step-by-step video tutorials, four coaching calls a week and 3,400+ members building real automations.
Is Hermes advanced delegation worth it?
That depends on how your Agent OS is set up today.
| Your Agent OS today | Is the plugin worth it? | Why |
|---|---|---|
| One Hermes profile, light delegation | Not yet. | A single delegation.model pin is simpler. |
| One profile, batches that mix easy and hard work | Yes, with model routes. | Each task picks a cheap or strong route. |
| Several specialist profiles with their own skills | Yes, with profile routes. | Tasks reuse each profile's model, skills and SOUL.md. |
| Work that must stay on your hardware | Yes, with a local model route. | A base_url route keeps that work local. |
| Durable jobs that must survive restarts | Use kanban instead. | Kanban tasks are durable and have their own per-task model override. |
My honest view is that most people should start with the single delegation.model pin, which I cover in the sibling post on the Hermes subagent model.
The plugin becomes worth it the moment your batches mix easy and hard work, or you already run specialist profiles.
How it fits an Agent OS
My Agent OS runs Hermes alongside other agents, and each job tends to live in its own Hermes profile.
That makes profile routes a natural fit, because the profiles already exist.
A main Hermes profile can hand coding tasks to a coder profile and research tasks to a researcher profile in the same batch.
Each subagent then runs as if you'd started hermes -p with that profile for the task.
For anyone running one profile, model routes give the same cost control with no extra profiles to maintain.
My Hermes kanban guide covers the board option for durable, named-profile work.
What the agent actually sees
Once a mode is active, each task item in delegate_task gains one extra property.
It is called option for model routes and profile for profile routes.
That property is a fixed list of your route names in config order, and its description lists each name with the one-line description you wrote.
Hermes refuses any other value before a subagent starts, so the agent can only choose from routes you approved.
A task that picks nothing runs on the mode's default route, or on the parent's own model when you didn't set a default.
With nothing configured, the property doesn't appear at all.
That's why route descriptions matter so much, because they're the only thing the agent reads when it decides where a task goes.
If the agent keeps choosing badly, rewrite the descriptions before you change anything else.
Failures, fallbacks and what the results tell you
Failures are explicit, so a missing or disabled backend returns a tool error naming the fix and no subagent starts.
Inside a batch, a task that can't be routed is refused on its own with route.refused: true, while the others run.
A model route without its own fallback_providers uses delegation.fallback_providers, and a profile route uses that profile's own chain.
Each result's route field shows the route name, provider, model, reasoning effort and speed tier the child actually used.
While either route mode is active, delegation.model and delegation.provider aren't used.
The limits to know
- It needs Hermes 0.22.5 or newer.
- Mode and route changes apply from the next session unless you use
--now. - Profile routes share the listed profile's credentials and terminal policy.
- Keys written into route config are stored in plain text.
- It doesn't make subagents durable, so a restart still doesn't resume a running child.
For model names to put in your routes, my best Hermes Agent model guide is the place to start.
If you want to talk through where routed agents fit your SEO business, you can book a free SEO strategy session with my team.
Also On Our Network
FAQ
What is Hermes advanced delegation?
It is an official Nous Research plugin that adds model routes and profile routes to delegate_task, so each subagent task can run on a named model or as one of your Hermes profiles.
How do I install it?
On Hermes 0.22.5 or newer, run hermes plugins install advanced-delegation, then hermes delegation config to pick a mode and build routes.
Is it worth it if I only use one profile?
Only if your batches mix easy and hard tasks. Otherwise a single delegation.model pin is simpler.
What's the difference between model routes and profile routes?
Model routes pick a provider and model per task, while profile routes run the subagent as a whole Hermes profile with its own model, credentials, skills and SOUL.md.
Why doesn't a mode change apply immediately?
The tool list is part of the prompt cache, so changes apply from the next session. /delegation <mode> --now applies it now at the cost of one cache break.
Is there a security risk with profile routes?
Yes. A listed profile's credentials, credential pool, SOUL.md and terminal policy become usable by the delegating profile's subagents, so only list profiles you trust with that access.
Final word on Hermes advanced delegation
Start with the single model pin, add model routes when your batches mix easy and hard work, and add profile routes once your specialist profiles are worth sharing.
If you run an agency and want help planning agent work for clients, you can book a free strategy session with my team.
If you want the whole Agent OS with the Hermes tab, the profiles and the training already set up, it lives inside the AI Profit Boardroom, with 3,400+ members and four coaching calls a week.
That is how I'd decide whether Hermes advanced delegation belongs in your setup.
About Julian
I'm Julian Goldie, an AI entrepreneur, SEO expert and the founder of the AI Profit Boardroom, which has 3,400+ members.
I help business owners scale with AI agents, automation and SEO.
My YouTube channel has 400,000+ subscribers, and I run Goldie Agency, a seven-figure SEO agency.
Related guides on agentos.guide
→ Hermes Kanban: The Self-Driving Board
→ Hermes Agent v0.22: The Quick Start
Get the whole Agent OS
The installable system behind every guide on this site — dashboard, agents, memory and pipelines — updated as new tools land, with 3,400+ members and four coaching calls a week.
Join the AI Profit Boardroom →📺 Video notes + links to the tools 👉
🎥 Learn how I make these videos 👉