2026-06-24

Hermes X Roundup: June 24, 2026 - Telegram Receipt-Split Bot, Slack-Fronted Agency Stack, /learn for Directories

hermesrounduptelegramslacklearncommunityx

YanXbt (YanXbt) shipped a working example of Hermes Agent as a Telegram-only utility on June 24. The user snaps a photo of a restaurant receipt, sends it to a Hermes bot in Telegram, and the agent reads every line, computes the split, and returns a message ready to forward to the people at the table. The tweet landed 8 likes, 1 retweet, 8 bookmarks, and 997 impressions in the first 90 minutes - small absolute numbers, but the engagement quality matters: 8 bookmarks on a 997-impression post is a 0.8% save rate, which is the threshold at which posts get recommended beyond the original author's followers. The quote-tweet attaches YanXbt's 15 Levels of Hermes Agent article, which has 388 likes and 179,950 impressions - the highest-engagement piece of long-form Hermes writing this week.

HERMES AGENT SPLITS YOUR RESTAURANT BILLS.
SEND A PHOTO OF THE RECEIPT ON TELEGRAM.
IT READS EVERY LINE, CALCULATES THE SPLIT,
AND GIVES YOU A MESSAGE TO SEND YOUR FRIENDS.

the workflow:

1. snap a photo of the receipt
2. send it to your Hermes bot on Telegram
3. agent extracts… https://t.co/KVMb2rppJW pic.twitter.com/w4S4CBPgA8
— YanXbt (@IBuzovskyi) June 24, 2026

Jess @ FireTeam documented a production tech stack on June 24 where Slack is the Hermes Agent frontend. The stack is Slack for comms and chat-with-Hermes, Fibery for project management, Claude as the model, plus Hermes Agent on top - Jess describes Slack as "obvious, but also frontend for our hermes agent." 9 likes, 9 bookmarks, 752 impressions, and three replies from builders asking how Slack-as-frontend is wired up. The configuration pattern is becoming a small genre: community members posting the same "here is how Hermes fits into my existing SaaS graph" diagrams.

Since y'all asked, here is the tech stack of just the production side of our creative agency...

✅ Slack - comms obvious, but also frontend for our hermes agent (see below)

✅ Fibery - PMS, basically a more flexible Notion. Automation heavy.

✅ Claude - you know this one.

✅… https://t.co/TvO5QISNDQ
— Jess @ FireTeam 🔥 An agency on fire... (@HireFireTeam) June 24, 2026

Cyril Gupta posted the most precise technical claim of the day on June 24: Hermes Agent's /learn command now accepts raw directories - codebases, API docs, configs, manuals - and distills them into a verifiable knowledge structure, bypassing the token-heavy context stuffing that pattern-matched retrieval would need. 2 likes, 18 impressions on the tweet itself is low, but Cyril's audience has been writing about Hermes in long form all month, and the technical frame "structured architecture is the leverage play over blind prompting" is the one builders are starting to repeat back. The /learn skill is documented on the Cat Labs blog in the June 24 /learn distillation post, which landed the same day.

Moving from blind prompting to structured architecture is the ultimate leverage play. Hermes Agent now lets you run /learn on raw directories—codebases, API docs, configs, or manuals.

It bypasses the token-heavy context stuffing and distills everything into a verifiable,… pic.twitter.com/04QKIXIKoN
— Cyril Gupta (@cyrilgupta) June 24, 2026

The highest-leverage case study of the day came from Jared Lucas, who stood up a managed Hermes Agent on his own infrastructure for a real estate client and connected it to Google Workspace and QuickBooks. The agent now handles the repetitive work that an executive assistant used to do - calendar, invoicing follow-ups, document prep. Jared reports the client dropped $18k per year in accounting fees and removed the executive assistant role entirely. 0 likes and 93 impressions is small, but the post is the cleanest before/after ROI this roundup has seen.

Helped a real estate agent drop $18k a year in accounting fees and eliminate the need for an executive assistant.

Stood up a managed @NousResearch Hermes agent on my infrastructure, connected it to his Google Workspace and QuickBooks. The agent now handles the repetitive work…
— Jared Lucas (@jaredclucas) June 24, 2026

Dale Thomas made the cleanest endorsement of the day on June 24: "I used to think @ManusAI was the G.O.A.T. now, I am convinced Hermes Agent from @NousResearch is. Running with GLM 5.2, it's unstoppable." 2 likes, 9 impressions - small engagement, but the post is signal that the provider-flexibility feature is paying off: GLM 5.2 is not a Nous-curated default, and a builder publicly picking it shows that the multi-provider story is real.

I used to think @ManusAI was the G.O.A.T. now, I am convinced Hermes Agent from @NousResearch is. Running with GLM 5.2, it's unstoppable.
— Dale Thomas (@dalethomas50) June 24, 2026

Notable Mentions

AlphaOracle is two weeks into Hermes Agent paired with the Obsidian Vault plugin - the central markdown repository is the value even for non-developers.
Kyz showcased Hermes Workspace (5.8k GitHub stars) as the dedicated web workspace built for Hermes Agent.
Stein shipped four production use cases on Hermes after months of setup - posting it as a question thread for comments.
Jared Lucas followed up with a second example: AI Slack Agents running daily reports on GA4 data.
Kenta.Kase reported a 400 error when using GitHub Copilot as a provider in Hermes - possibly limited to Business/Enterprise plans.

The thread across all five featured posts is the same: Hermes Agent is moving out of the "chatbot with skills" frame and into operational infrastructure. Telegram receipt-splitting, Slack-fronted creative agency stacks, Google Workspace + QuickBooks back-office automation, and 5.8k-star web workspace wrappers are all deployments where the agent runs on someone else's surface and produces outputs the user doesn't have to touch. The most-shared post of the day - YanXbt's 388-like article - frames the trajectory explicitly: "15 levels of Hermes Agent. From chatbot to 24/7 autonomous system."