Hermes Agent usage limits now work on daily and weekly windows, not one monthly block of credits. Nous Research announced the change on X on 27 October 2026: every new Hermes Agent plan includes a cloud agent out of the box, and existing subscribers can switch to the new plans any time on the Nous billing page at portal.nousresearch.com/billing.
📺 Watch: NEW Hermes Cloud + Agent OS is Crazy!
🔥 Get the Agent OS as a free bonus: AI Profit Boardroom members get the full Agent OS zip, prompt libraries, daily tutorials and weekly live coaching calls. → Get inside
Hermes itself was ready for it. v0.22.5, published on GitHub on 23 October 2026, added support for these window-based plans, with plan gauges, limit cards and recovery when you switch models. This guide covers how the limits work, where you see them, what happens on the free tier, and how to keep working when you hit one.
What changed in the new Hermes Agent plans
The Nous announcement makes three points:
- Daily and weekly limits. Usage refills on a daily and a weekly window instead of one pot of credits that lasts the month.
- A cloud agent in every new plan. Hermes Cloud runs a hosted Hermes Agent for you, so it keeps working when your laptop is shut. The Hermes docs say you manage those instances from the agents page in the Nous Portal.
- Switching is optional. Existing subscribers can move to the new plans whenever they like.
What about prices and the exact allowance per window? Nous hasn't put them in the Hermes docs, and they can change, so see the Nous billing page for the current numbers before you pick a plan.
Why does the shape matter? With one monthly block, a heavy week could drain the whole month. Daily and weekly windows spread the allowance out. A big day can hit the daily cap, but the next window refills it. The trade-off is that you now plan around two clocks instead of one balance.
I run Hermes every day inside my Agent OS, with a profile per model, so limits shape how I route work. If you want that setup ready to go, the AI Profit Boardroom ($69/mo) gives you the Agent OS, a 30-day AI agent roadmap, daily tutorials as new tools ship, 4 coaching calls a week and the prompt library. Over 3,400 members are in there.
📺 Watch: Hermes3D Is INSANE: AI Agents Now Have a 3D Office
How Hermes Agent usage limits show up in the app
The v0.22.5 release notes list "usage-window plan support" with three parts. Here's what each one is for.
| Feature | Added in | What it does |
|---|---|---|
| Plan gauges | v0.22.5 | Gauges for plans that run on usage windows |
| Limit cards | v0.22.5 | Cards for when a window-based plan reaches its limit |
| Model-switch recovery | v0.22.5 | Recovery when you switch models on a window-based plan |
| "% left · resets" chip | v0.22.0 | Shows a subscription running low, and when it resets, right in the model picker |
| Usage-limit cards with reset time | v0.22.0 | Desktop cards that show the reset time and can schedule a retry |
The v0.22.5 notes give these as a one-line summary. Nous says the full curated notes for this window ship with v0.23.0, so expect more detail then.
The model picker shows what's left
This one landed in v0.22.0. The model picker in hermes model, in /model and in Desktop shows a chip like "17% left · resets 3:00 PM" when a subscription runs low, and shows when a rate limit lifts. You can also star favourite models and pick reasoning effort in the same step. The point, in the release notes' own words, is that you switch before you hit the wall instead of after.
Check your limits on demand
You don't have to wait for a warning. These commands are in the Hermes docs:
- /usage shows token use, estimated cost and session length, plus an Account limits section with remaining quota, credits or plan usage pulled live from your provider.
- hermes usage shows your account's rate-limit windows from the terminal without opening a session. Add --json for scripts.
- /subscription (alias /upgrade) shows your Nous plan and opens the browser to change it. It's CLI only.
- /topup shows your Nous balance and handles billing on the portal. It replaced the old /credits and /billing commands.
- hermes portal info shows your login status, subscription info and how your model and tools are routed.
Status-bar notices for Nous credits are on by default. If they bug you, set display.credits_notices to false. /usage keeps working either way. For the rest of the commands, see my Hermes Agent slash commands guide.
Not sure which plan fits how you work? Book a free AI strategy session and we'll work it out.
📺 Watch: Hermes AI Agents Just Went Portable
The free tier's rate limit
Not paying yet? The Nous free tier has its own limit, and v0.22.0 changed how Hermes handles it.
- The turn ends at once. When you hit the free rate limit, Hermes stops the turn straight away and tells you when the limit resets, instead of sitting there retrying.
- Every failure explains itself. Each free-tier failure has a code and a plain sentence.
- A free model to start. A first desktop install with no provider and no Nous account gets the free nous/welcome model as a one-time trial, once per OS user. The CLI, TUI and gateway don't get a free model.
- Upgrade without losing your apps. hermes auth upgrade, or /login in any chat or messaging DM, turns the free identity into a full Nous account with your connected apps.
The configuration docs explain the logic. On the free tier, a session someone is watching (Desktop, the TUI, the hermes terminal chat or an editor over ACP) won't sit through a cooldown longer than 60 seconds. The pause covers the whole service, so a retry before the reset just meets the same refusal. If you've set a fallback, it runs first. Unattended runs, like hermes chat -q, cron jobs, messaging platforms and subagents, keep waiting.
What to do when you hit a limit
Hitting a window doesn't have to stop the work. Here's the order I'd go in.
- Switch model. Open /model and pick something with room left. The picker shows which subscriptions are running low, and v0.22.5 added recovery when you switch models on a window-based plan.
- Set up a fallback. Add a fallback_providers list to config.yaml. When the main model fails on a rate limit, Hermes tries the next provider. Fallback lasts one turn, and it's reset-aware: if the main provider has reported a reset time that hasn't passed, Hermes stays on the fallback until it does.
- Run a local model. Hermes can install llama.cpp and run a model on your own machine. No account, no limit window.
- Let unattended jobs wait. Rate limits that name a reset time get waited out, capped at 600 seconds per retry. For cron jobs on a subscription provider that reports its window exhausted, the docs say the scheduler can park the job until the window reopens instead of failing on every tick.
- Change plan. If you keep hitting the daily cap, look at the plans on the Nous billing page, or run /subscription.
One heads-up from the docs: switching model or falling back resets the prompt cache. The new model has to read your whole conversation again at full price, so don't bounce between models for no reason.
Where the cloud agent fits
The cloud agent is the other half of the announcement. Hermes Cloud runs hosted Hermes Agent instances that stay on around the clock. You manage them from the Nous Portal, and the Hermes docs also cover managing them from your local agent over MCP: list them, check status, start, stop and create new ones. My Hermes Cloud guide walks through the setup, and the video above shows it running with my Agent OS.
Coming from OpenClaw with a pile of provider keys? The Hermes docs suggest hermes setup --portal to collapse them into one Nous login. Do the import first with my guide on how to migrate OpenClaw to Hermes, then sort out the plan.
The plan you pick also decides which models you can reach, so check Goldie Bench for how they compare on real agent work. The Agent OS guide shows how I split work across profiles so one limit never stops everything. Want help with the setup? Grab a free AI strategy session.
FAQ
Do I have to switch to the new Hermes plans?
The Nous announcement says subscribers can switch to the new plans any time at portal.nousresearch.com/billing. It doesn't say the old plans are going away.
How much does a Hermes Agent plan cost?
Prices and allowances aren't in the Hermes docs. See the Nous billing page for current pricing before you choose.
How do I see how much usage I have left?
Run /usage in a chat, or hermes usage in the terminal. The model picker also shows a "% left · resets" chip when a subscription is running low.
What happens when the free tier runs out?
The turn ends straight away and Hermes tells you when the limit resets. A configured fallback runs first. Unattended runs like cron keep waiting.
Do I need Hermes v0.22.5 for the new plans?
v0.22.5 is the release that added plan gauges, limit cards and model-switch recovery for window-based plans. Update with hermes update, or through the Microsoft Store if you use the Store app.
Ready to make Hermes Agent usage limits work for you instead of against you? Join the AI Profit Boardroom for the Agent OS and daily Hermes tutorials, or book a free AI strategy session and we'll plan your setup together.
