Hermes Agent Update v0.9 is MASSIVE! (Everywhere Release)

v0.9 "Everywhere" ships 487 commits in 5 days

The release landed 5 days after v0.8 and is described as the biggest in project history. Hermes ships by commits rather than versioned drops — the CLI tells you how many commits you are behind on boot, and most of these features were already reachable via Hermes update over the weekend.

Local web dashboard on port 9119

You can now launch a browser UI to manage settings, sessions, skills, and the gateway without touching config files. By default it loads locally — do not expose it to the public, as the creators warn a 12 seconds SSH attack is realistic. If you're on a remote box, ask the agent to set it up. The channel has already migrated from OpenClaw to Hermes for this dashboard plus the self-improving skills loop.

Fast mode, background monitor, and backups

/fast mid-session routes to priority queues for GPT 5.4 coders and Quad models — same intelligence, roughly double the cost. Neither host uses it. The new background process monitor streams errors, port-listen messages, and terminal activity into chat, a deliberate design shift from OpenClaw's silent handling. Two CLI commands, Hermes backup and Hermes import, now snapshot and restore config, session, skills, and memory — replacing the host's old cron script.

Pluggable context engine and platform push

A new plugins command lets you swap in custom filtering, summarization, and injection logic. Honcho is still being tested for long-term memory. The "Everywhere" name signals a push into China via WeChat and iMessage support, and Hermes explicitly assumes everything is going to go to plan sideways — you'll still need external tools taped on top.

Update the right way

Run Hermes update from the CLI rather than asking the agent to update itself, then follow with the doctor command. Use the fleet commander if you manage multiple agents. Security hardening is still light — don't give it your whole life story.