Hermes Agent is the open-source, self-improving AI agent built by Nous Research. Bit-Sulting deploys it locally for your business — installed, integrated, hardened, and taught to do real work.
Not a coding copilot tethered to an IDE, and not a chatbot wrapper around a single API. Hermes is an autonomous agent that gets more capable the longer it runs. It lives wherever you put it — a spare mini PC, a $5 VPS, or a GPU box — and you talk to it from the apps you already use: Telegram, WhatsApp, Discord, Slack, or Teams, while it works on hardware you control.
Its defining feature is the built-in learning loop: Hermes creates skills from experience, improves them during use, nudges itself to persist what it learns, and builds a deepening model of who you are and how your business works — across sessions, across months.
Remembers your preferences, your customers, your conventions — injected into every session.
After a hard task it saves the procedure as a reusable skill — compatible with the open agentskills.io standard.
Telegram, WhatsApp, Discord, Slack, Teams and more. Your team messages the agent like a coworker.
Built-in cron: morning briefings, invoice chases, price watches, watchdog alerts — delivered anywhere.
Search, extract, drive a real browser, generate images, read documents — 60+ built-in tools.
Real-time voice interaction in the CLI and on Telegram/Discord. Hands-free ops.
Bot Mode: named specialist bots, each with its own model, memory and skills, coordinating in group chats.
Command approval, authorization policies, container isolation. You decide what it may touch.
Real workflows we deploy. Each one starts as a conversation and ends as a scheduled skill.
AI agents are becoming economic actors — paying for what they consume, collecting what they're owed, settling with each other. Bit-Sulting wires those rails into your deployment, from sats to stablecoins.
Each agent can carry its own Lightning wallet: receive payments, pay per-use APIs, data feeds and compute in sats, and settle with other agents — instantly, at near-zero cost. Micropayments finally make business sense.
Stablecoin flows through the processors, exchanges and on/off-ramp systems your business already trusts — fiat-stable value, programmable settlement, no new infrastructure to learn.
Per-transaction caps, daily budgets, human approval above thresholds, and an audit trail you can actually read. The agent spends, the math verifies, you set the limits — #BetOnMath applied to money.
| Rail | Best for | How it settles |
|---|---|---|
| Bitcoin / Lightning | Micropayments, pay-per-use APIs & data, agent-to-agent commerce | Sats — instant, near-zero fees |
| USDT / USDC | Invoicing, supplier payouts, international settlement | Stablecoins via your existing processors & on/off-ramps |
| Your fiat rails | Everything the business already does today | Your bank & processor, unchanged — the agent just does the work around them |
"It's not tied to your laptop. It lives where you put it — and it gets better at your business every week."— why we chose Hermes
| Option | What it is | Good for |
|---|---|---|
| Spare desktop / mini PC | Any x86 box in your office, repurposed as the agent host. | Zero new hardware cost. Our default recommendation. |
| NAS / home server | Docker deployment alongside your existing services. | Businesses that already run a NAS. |
| $5 VPS | Small cloud instance; agent reachable 24/7. | Always-on messaging automations, remote teams. |
| GPU box + local models | Local LLMs via Ollama/llama.cpp — zero API calls, fully private. | Strict data privacy needs; maximum sovereignty. |
| Cloud model APIs | Point the agent at frontier models when you want top-end reasoning. | Hard reasoning tasks; mix-and-match per bot. |
Most SMB deployments start on a repurposed box with a cloud model for reasoning and local guardrails — then migrate workloads local as needs grow. We design that path with you.
Every engagement starts with a free discovery call. Pricing is quoted per deployment — no mystery retainers, no per-seat fees to the software itself.
Hermes installed on your hardware or VPS, security hardening, one channel wired up (e.g. Telegram), and one real workflow automated end-to-end. Your team gets a hands-on walkthrough.
Multiple channels, scheduled automations, custom skills built from your actual processes, backups, monitoring, and training for everyone who'll talk to the agent. Payment rails — an agent Lightning wallet and stablecoin flows — included where they make sense.
Updates, health monitoring, new skills as your needs grow, priority troubleshooting. Your agent keeps compounding while we keep it healthy.
Prefer to try it yourself first? That's the spirit. Install it, or let us do it right the first time.
Your choice. With cloud model APIs, prompts go to the model provider — just like any AI assistant. With local models (Ollama/llama.cpp on your own GPU or CPU), everything stays on your hardware. We help you pick the right mix and label exactly what goes where.
The software is free — MIT licensed, no per-seat fees. Your costs are hardware you likely already own, any model API usage you choose, and Bit-Sulting's deployment/support fees, quoted up front after a discovery call.
Only what you allow. Hermes ships with command approval, authorization policies and container isolation, and we add firewall rules and least-privilege configuration as standard. Dangerous actions require your explicit sign-off.
Yes — and only as much as you allow. Payment rails go in as integrations: a Lightning wallet for sats and micropayments, USDT/USDC through payment systems you already use. Everything starts read-only (the agent drafts, you approve), and spend caps, approval thresholds and a full audit trail are in place before your agent pays anything on its own.
Telegram, Discord, Slack, WhatsApp, Microsoft Teams and more through the built-in messaging gateway. Your team messages the agent exactly like they message each other.
No. Hermes is open source, its skills are portable markdown (agentskills.io compatible), and your config and memory live on your machine. If you ever want to run it without us, everything is yours — we'll even document how.
We'll tell you on the discovery call. Agents are powerful but not magic — judgment-heavy decisions, legal commitments, and anything safety-critical stay with humans. The agent handles the grind; your people handle the calls.
Book a free discovery call — bring one repetitive task you hate, and we'll show you how the agent would take it.