Hermes advanced delegation is Nous Research's official plugin that lets Hermes Agent route each subagent task to a named model or to one of your Hermes profiles, and my verdict is that model routes are the upgrade most people should switch on first.

The plugin adds two modes to the delegate_task tool: model routes, called model_map, and profile routes, called profile_routes.

It is free, it needs Hermes 0.22.5 or newer, and it fixes the biggest limit of built-in delegation, which is one model for every subagent in a batch.

This review scores the plugin feature by feature and then puts model routes and profile routes head to head.

I haven't run a lab test of every route type, so the scores are my judgement of the documented behaviour on Nous's own plugin page and docs.

Hermes Advanced Delegation: The Scorecard

Here is the verdict on every feature up front.

Feature What it does Score
Model routes Each task picks a named provider and model. 9/10
Profile routes Each task runs as one of your Hermes profiles. 8/10
Hidden credentials Only names and descriptions reach the model. 10/10
Route reporting Each result shows the route, provider, model, effort and tier. 9/10
Explicit failures Missing backends and bad routes refuse loudly. 9/10
Fallback handling Routes use their own chain or delegation.fallback_providers. 8/10
hermes delegation config A guided editor that validates each route. 9/10
/delegation command Switch modes from any chat surface. 8/10
Session-scoped changes Changes apply from the next session to protect the prompt cache. 7/10
Install and version needs It needs Hermes 0.22.5 and installs per profile. 7/10

My overall rating for Hermes advanced delegation is 4.5 out of 5.

🔥 Want my exact Hermes Advanced Delegation Review setup, step by step? Inside the AI Profit Boardroom you get step-by-step video trainings, my installable Agent OS and four coaching calls a week with 3,400+ members building real automations. → Join the AI Profit Boardroom

What Hermes Advanced Delegation Is

Hermes Agent's main agent spawns subagents with delegate_task.

With the built-in default, every subagent runs on delegation.model and delegation.provider, or on the parent's model.

The delegation docs admit the catch, which is that the default pin is global and delegate_task has no per-task model parameter.

The advanced-delegation plugin is Nous's official answer, and it registers two delegation providers.

The plugin page says Hermes core keeps the whole subagent lifecycle, including spawning, isolation, depth and concurrency limits, background runs and results.

The plugin only answers one question, which is which route a task takes.

For how subagents work underneath, see my verdict on Hermes Agent subagents.

Model Routes (9/10)

Model routes let you define named options, each with a description, a model and a provider or direct endpoint.

The plugin page's example has three routes.

Route Description from the example Provider and model
fast Cheap and quick: lookups, summaries, simple edits. openrouter, google/gemini-3-flash-preview
deep Slow and careful: hard debugging and design work. anthropic, claude-opus-5, high reasoning effort
local Private: runs on this machine. A base_url on localhost with qwen3-coder

Each route can also take reasoning_effort, service_tier, fallback_providers, api_key and api_mode.

A route resolves exactly like the matching delegation.* keys, so a named provider uses the credentials you already configured.

It scores 9 because it gives you the cost control the built-in default can't, with very little setup.

It drops a point because an api_key written into a route sits in config.yaml in plain text, which the page itself warns about.

Profile Routes (8/10)

Profile routes let a subagent run as one of your Hermes profiles.

The plugin page says it runs exactly as if you'd started hermes -p <profile> for that task.

That brings the profile's own model, credentials, skills and SOUL.md along with it.

The list doubles as an allowlist, so a profile you don't list can never be picked, and a listed name with no matching profile is refused.

It scores 8 because it's the more powerful mode for specialist agents, but it carries the security catch I cover below.

Hidden Credentials (10/10)

This is the best design decision in the plugin.

Only the route names and one-line descriptions reach the model.

The raw provider, model id, endpoint and key never become tool parameters, so the agent can't invent a route or see a secret.

Hermes refuses any value outside your list before a subagent even starts.

Route Reporting (9/10)

Each routed result carries a route field.

It shows the route name plus the provider, model, reasoning effort and speed tier the subagent actually ran on.

The delegation docs add fallback_activated: true when a fallback provider took over.

That makes it easy to audit whether the agent is choosing sensible routes.

Explicit Failures (9/10)

The delegation docs say failures are explicit and never a silent switch.

If the selected backend is missing or disabled, delegate_task returns a tool error naming the backend and the fix, and no subagent starts.

Inside a batch, a task with an unknown option, a missing profile, an unsupported speed tier or missing credentials is refused on its own with route.refused: true.

The other tasks in the batch still run.

I rate this highly, because a surprise route is worse than a clear error.

🔥 Want my Hermes setups, scored and explained on video? Inside the AI Profit Boardroom, I've got Hermes Agent training, the Agent OS and walkthroughs on profiles, models and delegation. You'll be learning alongside 3,400+ members with four coaching calls every week. → Get access here

Fallback Handling (8/10)

A model route without its own fallback_providers uses delegation.fallback_providers.

Setting fallback_providers: [] on a route disables fallback for that route.

A profile route uses that profile's own fallback chain.

It scores 8 because the rules are clear, but you have to plan a chain per route if you want different backups.

The hermes delegation config Editor (9/10)

hermes delegation config opens a guided setup, and hermes delegation config --show prints the current mode and routes.

The model routes editor is built from hermes model's own pickers, so you choose providers you already have credentials for.

Each route is validated before it's saved, so the provider must be known, the effort must be a real level and the tier must be one the route offers.

Everything is written with Hermes's own config writer, which keeps your comments and other settings.

It scores 9 because it removes nearly every YAML mistake.

The /delegation Command (8/10)

/delegation shows the current mode and a picker, using buttons on Telegram, Discord and Slack and the clarify picker in the CLI, TUI and Desktop.

/delegation model_map saves model routes for the next session, and /delegation default switches back.

Mode words include models or 2 for model routes and profiles or 3 for profile routes.

It scores 8 because it works on every surface, though a mode with nothing set up is refused until you run the config editor.

Session-Scoped Changes (7/10)

Mode and route changes apply from the next session, which means /new.

The plugin page explains why, because the tool list is part of the prompt cache, and changing it mid-chat makes the next request re-read the whole conversation at full price.

--now applies a change to the current chat at the cost of one prompt-cache break.

Until you start a new session, a task that picks a route from the previous mode is refused with a message naming --now and /new.

It scores 7 because the reasoning is sound, but it trips people up the first time.

Install And Version Needs (7/10)

The plugin installs with hermes plugins install advanced-delegation, or from Settings → Model → Subagents → Install plugin in Desktop.

It needs Hermes 0.22.5, and the plugin page says 0.22.0 to 0.22.4 refuses to install or enable it.

Plugins install per profile, so it goes in the profile that delegates, while target profiles only need to exist with a working model.

It scores 7 because the version floor is recent, since 0.22.5 was published on GitHub on 23 October 2026.

Model Routes vs Profile Routes: Head To Head

Round Model routes Profile routes Winner
Setup effort A list of models in one profile. Separate working profiles first. Model routes
Cost control Each task picks a cheap or strong model. Each task gets the profile's model. Model routes
Specialist skills The subagent keeps the delegating profile's skills. The subagent brings the target profile's skills and SOUL.md. Profile routes
Credential exposure Credentials stay where they are. The target profile's credentials become usable. Model routes
Fallback The route's chain or delegation.fallback_providers. The target profile's own chain. Tie
Privacy A local route can run on your own machine. A local-model profile can do the same. Tie

Model routes win more rounds, and that's why I'd switch them on first.

Profile routes win the one round that matters for specialist teams, which is reusing a profile's own skills.

The Security Catch

The plugin page and the delegation docs both carry the same warning.

Listing a profile in profile routes lets the delegating profile's subagents use that profile's credentials, credential pool, SOUL.md and terminal policy.

So only list profiles whose keys and terminal access you're happy for the delegating agent to use.

If a routed profile runs shell commands, my Hermes Agent sandbox review covers the backends that keep those commands off your host.

Strengths And Weaknesses

These are the strengths.

  • It routes each task to the right model or profile in the same batch.
  • The agent never sees providers, model ids, endpoints or keys.
  • Results report the route each task actually used.
  • Bad routes and missing backends fail loudly instead of silently.

These are the weaknesses.

  • It needs Hermes 0.22.5 or newer.
  • Changes apply from the next session unless you accept a cache break.
  • Profile routes share the listed profile's credentials and terminal access.
  • Route keys written into config.yaml are stored in plain text.

If you'd like a second opinion on how agents fit your SEO business, you can book a free SEO strategy session with my team.

🔥 Want help deciding between model routes and profile routes? The AI Profit Boardroom has the Agent OS, Hermes training and four weekly coaching calls where you can ask me how to route your agents. → Join the AI Profit Boardroom

My video on the Hermes 0.22 release covers the release line this plugin sits on.

Related Reading

Also On Our Network

FAQ: Hermes Advanced Delegation

Is Hermes advanced delegation worth installing?

Yes, if your subagent batches mix simple and hard tasks, because each task can pick its own model or profile.

Which is better, model routes or profile routes?

Model routes suit most people, while profile routes suit teams that already run specialist profiles.

What are the downsides of Hermes advanced delegation?

It needs Hermes 0.22.5, changes apply from the next session, and profile routes share the listed profile's credentials.

Does the plugin replace delegation.model?

While either route mode is active, delegation.model and delegation.provider aren't used, and switching back to Default restores them.

What happens if the plugin is disabled?

Hermes refuses the subagent with an error naming /delegation, and switching to another mode is the fix.

If you run an agency and want help planning agent work for clients, you can book a free strategy session with my team.

📺 Video notes + links to the tools 👉

🎥 Learn how I make these videos 👉

🆓 Get a FREE AI Course + Community + 1,000 AI Agents 👉

About Julian

I'm Julian Goldie, an SEO entrepreneur, author and founder of the AI Profit Boardroom, which has 3,400+ members.

I help business owners scale with AI agents, automation and SEO.

  • I've built a 7-figure SEO and link-building agency, Goldie Agency, with 70+ people in Bangkok.
  • I've grown a YouTube channel to 400,000+ subscribers.
  • I wrote the book "Link Building Mastery", which covers nine link building strategies.
  • I run GoldieBench, where I score AI models on real one-shot builds.

→ Get my best AI training inside the AI Profit Boardroom

Start with model routes, add profile routes once your specialist profiles are worth sharing, and Hermes advanced delegation earns its 4.5 out of 5.