five new Hermes Agent features worth knowing about: (1) see what a paid MCP tool costs before it runs (#142470) an MCP server entry can declare what each tool costs and a spend limit per chat. the approval prompt shows the cost and the chat's running total, and a call that would pass the limit asks first. (2) the classic CLI has a new default look (#139241) every line in the scrollback starts with a single glyph, and the status line sits under the input box. (3) long MCP tool calls show their progress (#141985) when an MCP server reports progress, Hermes Desktop shows the percentage and message next to the running tool's timer. the CLI spinner shows the same text. (4) pick the model your subagents use from a dropdown (#141312) Hermes Desktop settings now have separate provider and model dropdowns for subagents. (5) drag a session to pin it (#142672) in the Hermes Desktop sidebar, drop a session on the Pinned section to pin it, and drag it back to the session list to unpin it.

when does "0 9 1 * 1" run? if you said the first Monday of the month, it runs on the 1st and on every Monday. crontab-ref is a community plugin that gives your Hermes Agent the plain-words reading of any cron expression in your message: what each field means, the next five run times, and whether /cron will take it. no API key needed

your Hermes Agent can message you first: when mail arrives, a file is shared or a build finishes. hookspool is an open-source community project: a local inbox in front of your Hermes webhook routes that keeps every event on disk until your Hermes Agent takes it. > it saves each webhook to disk before it answers the sender > it holds events while your gateway restarts and delivers them when it is back > it turns a burst of events into one wake-up > bobbins, one plugin per service: Gmail, Google Calendar, Google Drive, GitHub, Trello, Kaneo, Zoom and Plaid > one general bobbin for any service that sends signed JSON > setup steps in every bobbin, one for you and one for your agent > it works with the Hermes webhook routes you already have > plain Python and one SQLite file, MIT licence what you can do with it: > get a message on Telegram, Discord or Slack when a new email arrives, with a draft reply from your agent > get a heads-up when someone shares a file with you in Google Drive > get one message after a burst of CI runs > ask your agent for a meeting brief when a new meeting lands on your Google Calendar

the community added 144 plugins to the Hermes plugin catalog this week. the six I'd install first: > llmslim: compacts older tool output on your machine, with no model call to summarise it, so prompt caching keeps working > hermes-memory-pane: shows what your Hermes Agent saved to or forgot from memory in the current chat, and lets you edit or undo it (Hermes Desktop) > advisor: a second model reviews every turn and sends its findings back as advice > agent-station: a read-only Hermes Desktop page for the subagents your Hermes Agent starts, with a replay of file edits > ux3-product-design-harness: lets your Hermes Agent review a product decision before you build it and give you one verdict > hermes-pantheon: eighteen Hermes Desktop themes drawn as marble gods, switched with /skin <name> thanks to everyone building and sharing these!

planning a long weekend away usually starts with a pile of flight and hotel tabs. viajante is a community plugin for Hermes Agent that searches flights and hotels from your own machine, with no API keys and no account. everything it covers: - reads the public Google Flights and Google Hotels pages from your own IP address - returns prices, flight times and links - gives your agent 22 tools plus a skill that explains how to read the results - books nothing and stores no credentials what you can do with it: - ask for a round trip with your dates and currency, and get nonstop options with flight numbers and times - ask about both Paris airports and have it pick one flight and one hotel, with its reasons - find hotels for the same dates, priced for the whole stay, with a minimum rating

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.

I gave Qwen3.8 27B, Flint and Muse Glimmer 30B the same Mont-Saint-Michel voxel prompt on my RTX 5090, with one hour of /goal in Hermes Agent each: >Qwen built most of the scene: the abbey, the tide and a night with lit windows. >Flint built the mount and a rough abbey, with no causeway and a night too dark to see. >Muse finished first, in 21 minutes, with a cone-shaped mount, a spire and a night key that does nothing. one run each. Flint ran at its maker's settings, Muse on defaults since its GGUF ships none (its maker-settings rerun did worse).