Hermes Agent v0.22.0 lowers the door: no-key setup, app connectors and one-click local models
By AgentRiot Editorial
Nous Research’s Threshold release lets a new desktop user start without an API key, connects the agent to Gmail, Linear and Notion, and runs open models locally through llama.cpp. It also moves installs onto tested stable releases and changes several defaults existing users should check.

Nous Research published Hermes Agent v0.22.0, “The Threshold Release,” (opens in new tab) on October 13. It follows v0.21.0 Pantheon (opens in new tab) and rolls up the v0.21.1 to v0.21.7 patch releases. The project reports about 24,200 commits and 6,080 merged pull requests since v0.21.0, and calls it the largest window it has shipped. The theme is getting through the door: fewer things a new user has to set up before Hermes does anything useful.
A first run that needs no key
On a computer where Hermes has never been used, the desktop app now opens on a short setup chat instead of a settings screen. It asks for a name, an accent color and a layout, which apps to connect and which plugins to install, then offers up to four first tasks. That chat runs on a free trial model, nous/welcome, which the release notes (opens in new tab) limit to a one-time trial per operating-system user on a first desktop install. The CLI, TUI and messaging gateway get no free model. Skip setup ends it, and Settings can run it again.
Connectors carry the agent into hosted apps
Hermes can now act inside hosted apps such as Gmail, Linear and Notion through the Nous tool gateway. You ask for an app, sign in once in the browser, and the tools are usable in the same turn. No Nous account is needed to start: the first time you use an app, Hermes creates a free, connectors-only identity, which hermes auth upgrade or /login can later convert into a full account.
Those requests route through Nous’s hosted gateway rather than directly from your machine. The release notes list two switches for people who would rather not use that path: nous.guest: false turns off guest identities, and tools.connectors.enabled: false turns connectors off. A new Connectors page in the desktop app lists hosted apps and your MCP servers together, replacing the old MCP tab, and supports several named accounts per app, such as a work and a home Gmail.
Local models without an account
For people who would rather keep everything on their own hardware, pull request #100667 (opens in new tab) adds a managed llama.cpp runtime: Hermes installs and updates the engine for your hardware (CPU, CUDA, Metal, Vulkan or ROCm), rates catalog models against your GPU and RAM, and picks the best build that fits. Downloads pause and resume, and the server starts and stops with Hermes. The catalog includes Qwen3.8 27B, Qwen3.6 35B-A3B and DeepSeek V4 Flash, and any GGUF from Hugging Face can be added. It is under Settings, Providers, Local Models, or hermes model in a terminal.
Stable releases and signed desktop packages
Source installs now follow tested SemVer releases by default instead of every commit on main. Each release must pass CI, Docker, Nix, signed desktop builds and install and upgrade runs on every OS before it ships; v0.21.6 and v0.21.7 were the first through that gate. hermes update --set-channel main switches back to the bleeding edge, and moving to an older release is labelled a rollback and needs confirmation.
The desktop app is now a signed package that bundles its own Python and Node. Windows gets an MSIX for x64 and ARM64 through App Installer or the Microsoft Store, and macOS gets a signed, notarized Apple Silicon build. A new package manager, hermes pm, owns pinned, hash-checked tools and environments, and updates switch to a new environment only after it passes a check, so a killed update should no longer leave an install half-applied.
The Python package index is not a reliable install route for this release. PyPI (opens in new tab) still listed 0.19.0 as the latest hermes-agent version on October 14. Use the project’s installer, desktop packages or hermes update.
Also in the release
The browser agent can sign in with logins from 1Password, Bitwarden or Hermes’s own encrypted vault, with the project saying the model never sees the password and that card and address fills need confirmation. A Bot Screen lets Hermes Desktop stream a Linux gateway bot’s desktop so you can take over for a login, 2FA prompt or CAPTCHA. Running subagents are now visible with live transcripts and Steer and Stop controls. A plugin catalog launched with 600 entries pinned to reviewed commits, and memory providers such as Honcho, Mem0 and Supermemory moved into it as plugins. The project also reports cutting the codebase from about 1.06 million to 698,000 lines with no behavior change.
Defaults that change on upgrade
The upgrade notes (opens in new tab) include several that affect existing installs:
- Ended sessions inactive for 90 days are now deleted at startup, because
sessions.auto_prunedefaults to on. Pinned and open sessions are left alone; setsessions.auto_prune: falseto keep everything. - Chats no longer reset on an idle or daily timer. The old
session_resetmodes are retired, and/newor/resetstarts fresh. - One messaging gateway and one desktop backend now serve every profile on a machine.
hermes updatefolds per-profile gateways into the shared one when nothing conflicts, and each profile keeps its own credentials, allowlists, cron jobs and sessions. - Managed installs run on Python 3.14. The Windows desktop is MSIX-only and needs Windows 11 22H2 or newer, and default macOS builds are Apple Silicon only.
- Plugin authors lose the old compatibility import layer and
hermes sync, and several tools were renamed, though the old names still work.
If you keep session history you care about, set the auto-prune option before updating. After that, hermes update follows the stable channel by default.

