By Lee Jie Teh, Founder · BixTech
Hermes is Nous Research's self-improving open-source AI agent. It lives on your server, talks back through Telegram, WhatsApp, Slack, Discord or Signal, runs scheduled work in the background, and learns new skills as you use it. BixTech installs and tunes it for your business, so you get an autonomous assistant from day one.
RM 2,000 one-time setup · No monthly fees · Includes training
Book Free Discovery Call on WhatsAppHermes is Nous Research's open-source personal AI agent — the same lab behind the Hermes LLMs. It is positioned as "the agent that grows with you" rather than a coding copilot or browser-based chatbot.
Hermes maintains memory across every conversation and auto-generates new skills based on how you actually work. The longer you use it, the more accurate it gets — without re-prompting.
Natural-language cron means you can say "every weekday at 8am, summarise yesterday's enquiries and DM me the top 3" and Hermes runs it forever — no extra automation tool.
Hermes delegates work to isolated sub-agents running in one of six backends (local, Docker, SSH, Singularity, Modal, Daytona). Parallel research, browser tasks, or file work — without polluting your main session.
200+ models out of the box — OpenAI, Anthropic, Nous Portal, OpenRouter, NVIDIA NIM, Hugging Face, and custom endpoints. Swap models per skill, per channel, or globally.
BixTech tunes Hermes to your role and daily workflows — scheduled and sub-agent automations land in your messenger from day one.
Hermes runs an 8am scheduled briefing: yesterday's sales, top enquiries, calendar conflicts, and any flagged emails — delivered to Telegram before your first meeting. Sub-agents handle deeper research in parallel ("compare these 3 competitors and DM me a one-page summary by 5pm") while you stay in your existing chat.
Hermes drafts captions, repurposes long-form into social, and monitors competitor mentions on a schedule. The learning loop means the more you accept or reject suggestions, the closer the output gets to your house style without re-prompting.
Auto-tunes to your house style over timeHermes spawns sub-agents to research topic angles in parallel, drafts outlines, repurposes long-form articles into social posts, and proofreads in multiple languages. Sub-agent isolation means one runaway research task does not crash your main editing session.
Parallel research with sub-agentsHermes watches your inbox, qualifies incoming enquiries, drafts personalised follow-ups, and runs end-of-day pipeline summaries — all through Telegram or WhatsApp. Persistent memory means it remembers each prospect's history across weeks, not just within a single chat.
Remembers every prospect across sessionsHermes processes invoice attachments inside a sandboxed sub-agent, extracts line items, flags discrepancies, and posts a daily payables summary to Slack. Sandbox isolation means PDF parsing or OCR cannot reach your main filesystem.
Sandbox-isolated invoice processingHermes answers staff questions from your SOP library, drafts internal announcements, and runs scheduled checks (e.g. weekly contract-renewal reminders). The learning loop captures recurring staff questions and surfaces them as candidates for new SOP entries.
One-time fee. No monthly retainer. No hidden costs. You own and operate your AI agent independently after setup.
Prefer to set it up yourself? Read our free install guide for Mac mini & VPS — and WhatsApp us if you get stuck.
We learn about your role, the scheduled or sub-agent workflows you want, and the messenger you actually live in — then map out exactly which Hermes skills to ship in week one. No commitment required.
BixTech installs Hermes on your Mac mini, Linux VPS (Singapore region for low MY latency), or existing server, sets up the daemon so it survives reboots, and runs hermes doctor to confirm a clean baseline.
We connect your preferred LLM provider — Claude, ChatGPT, Nous Portal (one-click via hermes setup --portal), OpenRouter (for 200+ models including Llama and DeepSeek), or a self-hosted endpoint — and set per-skill model routing.
Hermes is connected to Telegram, WhatsApp, Slack, Discord, or Signal via hermes gateway setup, then locked down: DM allowlist on your number(s), a sandbox backend selected for sub-agents (Docker for most SMEs, Modal for serverless), and the gateway bound to loopback or a Tailscale tailnet.
We tune the workflows identified in discovery — scheduled briefings, sub-agent research, inbox triage, document summarisation, SOP Q&A, or whatever your role needs. Skills are seeded so the learning loop has a strong starting point.
Live walkthrough — daily usage on your messenger, how the scheduler reads natural language, how to add or revoke skills, when to spawn sub-agents, and how to read the hermes doctor output yourself.
Two weeks of WhatsApp support for any questions, scheduled-skill tweaks, sub-agent issues, or additional workflows. After that, you're fully independent — with optional ad-hoc support available.
BixTech has delivered over 80 AI and software projects since 2018 for clients including DBS Bank, Digi, and Malaysian government ministries. We have set up both OpenClaw and Hermes deployments and know which tool fits which team — no upsell to the more expensive one.
We configure Hermes for Bahasa Malaysia, Mandarin, and mixed-language use, point model traffic at Singapore-region endpoints where possible to keep latency under 250ms, and integrate the local tools you already run (SQL Account, AutoCount, WhatsApp Business).
RM 2,000 once. You own everything: the Hermes install, the config, the skills, the chat history. If you outgrow the basic setup and want custom AI development beyond a personal agent, BixTech offers that too — but this package is designed to be complete on its own.
Hermes is a free, open-source AI agent built by Nous Research — the same lab behind the Hermes large language models. It runs on your own server, talks back through messengers like Telegram, WhatsApp, Slack, Discord, and Signal, remembers your work across sessions, and auto-generates its own skills based on how you use it. It is released under the MIT license and supports 200+ LLMs from OpenAI, Anthropic, Nous Portal, OpenRouter, and others.
Both are self-hosted, open-source personal AI assistants. OpenClaw is Node.js-based with a polished macOS / iOS / Android companion app, voice (Wake + Talk Mode), and 20+ messenger integrations — best for individuals who want a desktop-grade personal AI. Hermes is Python-based and CLI/server-first, with a self-improving learning loop, persistent memory, parallel sub-agent delegation, six sandbox backends, and a built-in scheduler — best for businesses that want autonomous, always-on AI doing real work in the background. See the full Hermes vs OpenClaw comparison.
BixTech offers a one-time Hermes setup package for RM 2,000. This includes a discovery call, full server install, model provider configuration, messaging gateway setup, sandbox hardening, 3–5 tuned workflows or scheduled skills, and a hands-on training session. No monthly retainer required — ongoing cost is the LLM provider you choose (Nous Portal flat rate, or pay-as-you-go via OpenAI / Anthropic / OpenRouter).
No. Once BixTech sets it up, you interact with Hermes by sending text messages on Telegram, WhatsApp, or your preferred messenger — the same way you already chat with colleagues. You do not need to know Python, run commands, or read logs. If you are technical and prefer to install it yourself, see our free step-by-step install guide for Mac mini & VPS.
Hermes runs on your own server, so chat history, memory, and skill state stay local. Each sub-agent runs inside an isolated sandbox (local, Docker, SSH, Singularity, Modal, or Daytona) so risky tool calls cannot touch your main system. Queries are still sent to whichever model provider you choose; for sensitive data, BixTech can route to a self-hosted model via OpenRouter or a local endpoint, keeping prompts and responses inside Malaysia.
Yes. Hermes is model-agnostic — it works with any of the 200+ LLMs accessible through OpenAI, Anthropic, Nous Portal, OpenRouter, Hugging Face, and custom endpoints. All major frontier models support Bahasa Malaysia, Mandarin, Cantonese, Tamil, and code-switched English/BM. BixTech configures the Hermes system prompts and skill descriptions in your preferred language during setup.
Yes. Hermes ships with a built-in OpenClaw import command (hermes claw migrate) that pulls your existing skills, channel connections, and history across. BixTech can either run the migration for you as a one-off, or include it inside a fresh RM 2,000 Hermes setup that also adds the Hermes-only features (sub-agents, scheduled cron, learning loop).
The typical Hermes setup is completed in 1–2 sessions totalling 3–4 hours of our time (longer than OpenClaw because of sandbox configuration and scheduled-skill seeding). You'll be using Hermes independently within 3–5 business days of the discovery call.
See the AI projects BixTech has built for clients
From AI editorial assistants (TanTanNews) to smart quotation engines (OMNE+) and autonomous sales systems (PropertyLab) — BixTech also builds fully custom AI automation for businesses that need more than an off-the-shelf Hermes setup.
View AI case studies & portfolioNot quite the right fit? Here are other ways we help Malaysian SMEs automate with AI.
Book a free 30-minute discovery call. We'll map out the best Hermes scheduled and sub-agent workflows for your role and business — no commitment, no hard sell.
RM 2,000
One-time setup · No monthly fees · 14-day support included
Or email contact@bixtech.co · Call +60 11-1300 4052