g-agent
galyarderlabs/galyarder-agent
Experimental local-first agent runtime for persistent characters, derived from nanobot and Hermes Agent
g-agent is an experimental local-first agent runtime derived from nanobot and Hermes Agent. It keeps persistent memory and works across WhatsApp, Telegram, Discord, Email and a CLI.
What g-agent does
g-agent lets you run agents that carry persistent memory, domain expertise and custom tools, and treats them as departments such as finance, devops, content or customer support. The same agent can respond over WhatsApp, Telegram, Discord, Email or the command line, and it can work with Gmail and Calendar through the local gws CLI.
The README lists what exists today: company identity and goals kept in local workspace files and SQLite, secure local tools for file editing, shell execution and task schedules, multi-model routing through direct providers, local inference or OpenAI-compatible proxies, and durable memory in MEMORY.md, PROFILE.md, PROJECTS.md and LESSONS.md. Scheduled jobs, cron triggers, event listeners and user-level systemd daemons are included. Its stated principles favour local control, explicit permission gates, approvals, allowlists and auditable action logs. Setup is a git clone followed by a pip editable install, and fuller guides are on its docs site. The license is MIT.
Key features
- Persistent memory in MEMORY.md, PROFILE.md, PROJECTS.md and LESSONS.md
- WhatsApp, Telegram, Discord, Email and CLI channels
- Google Workspace access through gws
- Local tools for files, shell and task schedules
- Multi-model routing including local inference
- Cron triggers, event listeners and systemd daemons
When to use it
- Running a personal operator that manages email and calendar
- Splitting work into finance, devops or support agent departments
- Keeping agent state on your own machine across restarts
Who it is for: Builders who want a self-hosted agent runtime with persistent memory and multi-channel presence.
How it fits with Hermes Agent
The repository states that it is derived from nanobot and Hermes Agent, making it a derivative runtime rather than a Hermes plugin.
How to install g-agent
These commands are copied from the project's README. Check the repository for the latest steps before you run them.
git clone https://github.com/galyarderlabs/galyarder-agent.git
cd galyarder-agent/backend/agent
pip install -e .
g-agent onboardNote: The repository describes itself as an experimental runtime.
FAQ
What is g-agent?
g-agent is an open-source, local-first agent runtime with persistent memory, channel presence and tool access across WhatsApp, Telegram, Discord, Email and CLI.
Does g-agent work with Hermes Agent?
It is derived from nanobot and Hermes Agent rather than being a plugin for it. The repository does not describe a way to run it inside Hermes Agent.
How do I install g-agent?
Clone the repository, change into backend/agent, run pip install -e . and then g-agent onboard. The README also lists g-agent status and g-agent gateway.
Is g-agent free and open source?
Yes. The repository is published under the MIT license.
Similar automations for Hermes Agent
All automationsLocal AI investment assistant with multiple agent engines, including a Hermes engine
mantenasud Hermes Notion GBrainHermes skill that turns WhatsApp or Slack messages into Notion tasks and syncs weekly summaries to GBrain
antiwork GumclawRedacted showcase of the Hermes Agent setup behind Gumroad operations, with 273 skills and 78 cron loops
AtlasOmnia Hermes AutoresearchArchived harness that loops an agent over a Git repo, keeping changes that raise a score
SouthpawIN KlerikHermes Agent profile that reviews other agent profiles' work against quality standards
AzamatSafarov Hermes NotebookLM BridgePipeline that turns NotebookLM audio, infographics and slides into platform-ready publishing assets
Related guides: Build agent teams with the Hermes kanban board · How to run Hermes Agent securely