
If you run Hermes Agent, start here. Commands you can paste, Setups you can copy and doc for your setup. Unofficial Tips for @NousResearch Hermes Agent
United States
Joined February 2026
- Tweets11.5K
- Following1.2K
- Followers11.2K
- Likes17.7K
Pinned Tweet
your Hermes can now literally call your iPhone... and hear your answer
it rings with a real incoming-call screen, speaks its question out loud and what you say goes back to Hermes as text
the rest of the time you just text it from the /call-me iPhone app like any other chat, and your scheduled jobs can text you there too
only the numbers you list can text it or get called, and messages have to come from that number's verified device
it's a community plugin called call-me, in the Hermes plugin catalog
sick apparel 🪽👀
ngl @NousResearch has some of the best looking merch out there. straight gas
CLI commands:
hermes plugins install call-me --enable
hermes gateway restart
when the install asks for CALLME_ALLOWED_USERS paste the 10-digit number your /call-me app shows (comma-separate more than one)
first time running the gateway? do hermes gateway install before the restart
treat those numbers like a secret... anyone who has one can text or ring that phone through the service
heads up... the app is a paid subscription (the plugin itself is free) and calls go through the author's /call-me service, which transcribes the audio and keeps the texts and transcripts
Needs Hermes 0.21.5 or newer.
Plugin: github.com/radres/hermes-plu…
App: callmemcp.com/aiphone
Docs: hermes-agent.nousresearch.co…
your Hermes is probably summarizing your long chats with your main model... yep, the pricey one
Hermes out of the box most of its side jobs run on the model you picked in hermes model... compression, session titles and more
the docs even say that on expensive reasoning models these side jobs add meaningful cost
you can move just compression to a cheaper model and then your main chat can stay the same
you can also turn its thinking down to low so the summaries don't overthink
no plugin needed, it's built in
CLI commands:
hermes config set auxiliary.compression.provider openrouter
hermes config set auxiliary.compression.model google/gemini-2.5-flash
hermes config set auxiliary.compression.reasoning_effort low
in your Desktop app: Settings > Model > Auxiliary models > Compression > Change, then pick a provider and model, set Reasoning to Low and hit Apply
Needs Hermes 0.19.0 or newer (0.21.3+ for the Reasoning picker in Desktop).
Docs: hermes-agent.nousresearch.co…
what's the cheapest model that's ever done a real good job in your Hermes… without you noticing the difference?
name it, taking notes
my Hermes Agent now sends each subagent to the model that fits the job.
I set up the new advanced-delegation plugin from Nous Research with two routes on Nous Portal, then gave it one prompt with two jobs:
- a quick summary of Python's heapq went to Claude Haiku 5.5 and came back in 3.0s
- a five-pirate logic puzzle went to Claude Opus 5.5 at high reasoning and came back solved in 9.3s
the main agent picked the route for each job by itself.
the whole run is in the video, setup in the first comment.
where does your Hermes actually live day to day…?
telegram, discord, the terminal, desktop, or a machine you forgot about
one web page is all it takes to hand your Hermes new orders...
Airlock makes the risky stuff wait for your yes... and once a session has read something from outside, even more waits for you
it's a community plugin, new in the Hermes plugin catalog... and it runs fully local, no network or LLM calls
web pages, MCP tool output, customer messages if you turn that on... any of those flips the session
messages to anyone but you, pushes and deploys always stop at Hermes's approval prompt... after the flip, memory and skill writes and browser clicks do too
reading, searching and drafting stay free
it even refuses sneaky commands like base64 | sh until they're written in plain sight... pretty cool right
finally we get an extra layer against prompt injection!
to use it... do
hermes plugins install airlock --enable
hermes gateway restart (only if a gateway is running)
hermes airlock demo
then start a new Hermes session so it loads
the demo doesn't run anything for real
with --yolo, approvals.mode: off or cron set to approve, it blocks those actions instead of auto-approving... set approvals_off: follow in ~/.hermes/airlock.yaml to get Hermes's normal behavior back
Needs Hermes 0.21.6 or newer.
Plugin: github.com/otsoa-dubief/airl…](github.com/otsoa-dubief/airl…
Catalog:
hermes-agent.nousresearch.co…](hermes-agent.nousresearch.co…
this is exactly my thoughts... there is no better way to put it and @jumperz did just that
I think is more than just whos better than who or what's better than what... gotta zoom out sometimes and look at the bigger picture
I’m still getting “why use hermes when grok bot exists”
so here’s why
grok bot has really good UX.. they really understood what the mainstream user wants and got them into agents so easily..
but it’s a managed agent, xai’s harness, xai’s models, xai’s rules..
hermes is a harness you own.. everything around the model, the loop, tools, memory, context, where the code runs..
the model thinks, the harness decides what it remembers, what it can touch and how it recovers when something breaks.. and with hermes the harness is yours and the model is swappable..
you can use them independently, but you can also wire them into a loop through a shared channel, where grok bot catches what’s moving on X, hermes runs it through its tools, memory and skills, then grok bot checks the output against what’s live..
seriously, you would be unstoppable
I think more and more people are going to start to realize this! glad Denis did... welcome to Hermes Agent sir!
I liked OpenCode, but they no longer deserve my trust 😔
As a security professional, I reached out to them several times over security and privacy issues I found.
I never heard anything back 🙈
Their poor handling of the data breach this week was the last drop, I'm afraid.
I really hope OpenCode gets better, but as it stands right now, their transparency track record leaves much to be desired.
So I'm planning to migrate to Hermes, since Nous actually seems to have security-minded folks on their team, who approach it the right way.
If you could only run one model in your Hermes for a month… which one are you picking? no money limits
okay this one is sneaky good...
one Hermes gateway can run a different model in every Discord channel
cheap and quick in #daily... your best model plus a code-review prompt in #dev
it's built in, it's called channel_overrides... no plugin needed!
anything you leave out falls back to your normal config and threads inherit their parent channel automatically
the channel prompt replaces your global gateway prompt for that channel only
One server, a whole team of Hermes!
set this up by adding this to ~/.hermes/config.yaml
platforms:
discord:
channel_overrides:
"123456789012345678":
model: openai/gpt-5-mini
"987654321098765432":
model: anthropic/claude-sonnet-4.6
provider: anthropic
system_prompt: "You are the #dev channel code-review specialist."
the model names are just examples... with no provider set, the model has to be on your default provider
swap in your own channel IDs (turn on Developer Mode in Discord under User Settings > Advanced, right-click the channel, Copy Channel ID)
then restart the gateway
hermes gateway restart
Needs Hermes 0.18.1 or newer (0.21.6+ fixes override channels rebuilding the agent every turn)
Docs: hermes-agent.nousresearch.co…









