Which computer should run your agent?
OpenClaw's hardware requirements are small: 1 GB of memory, or 4 GB for Claude Code and Codex, with the model in the cloud. Start on a computer you have. Buy a Mac for iMessage, and 32 GB for big local models.
Every pick, by price
The maker's list price, checked 18 October 2026. The Amazon buttons open the live listing.
Free: use what you have
The computer you already haveFreeFine for any of the four agents with a cloud model, if you set it never to sleep. Any 64-bit computer with 4 GB of memory, 8 GB to be comfortable.Illustrative image: a Dell OptiPlex Micro mounted behind a monitor.
Under $400: a small box for a cloud model
Dell OptiPlex 7070 Micro, refurbishedabout $237Cloud-model agent, always on, cheapest box with 16 GB that doesn't feel slow. Intel Core i5-9500T (6-core), 16 GB memory, 256 GB NVMe SSD.Illustrative image: a Dell OptiPlex Micro, an office-surplus mini PC.
GMKtec G3S$249.99Cloud-model agent on a new, tiny, low-power box at the lowest price; 8 GB meets Codex's recommended RAM. Intel N95 (4-core), 8 GB memory, 256 GB SSD.Illustrative image: an Intel NUC mini PC held in one hand.
Beelink EQ14$385Cloud-model agent on a new, quiet N150 box with 16 GB and a built-in power supply; the popular homelab pick. Intel N150 (4-core), 16 GB memory, 500 GB SSD.Illustrative image: a mini PC (an older Zotac ZBOX) showing its ports.
$600 to $1,300: room for small local models
GMKtec M6 Ultra, 32 GB$679.99Cheapest new 32 GB mini PC for a busy Linux agent host. AMD Ryzen 5 7640HS (6-core), 32 GB memory, 1 TB SSD.Illustrative image: a mini PC (an older Zotac ZBOX) showing its ports.
Beelink SER8, 32 GB$859Linux/Windows agent box with 32 GB for many tools, containers and browser sessions at once. AMD Ryzen 7 8745HS (8-core), Radeon 780M, 32 GB memory, 1 TB NVMe SSD.Illustrative image: a Lenovo ThinkCentre Tiny beside a coffee mug, for scale.
Mac mini M6, 16 GB$899Always-on cloud-model agent on macOS (needed for OpenClaw's iMessage), silent and 4 W at idle. Apple M6, 12-core CPU, 12-core GPU, 153 GB/s, 16 GB memory, 256 GB SSD.Illustrative image: a Mac mini (2024 design).
Mac mini M6, 24 GB$1,299macOS agent host with headroom for browser automation and mid-size local models. Apple M6, 12-core CPU, 12-core GPU, 170 GB/s, 24 GB memory, 512 GB SSD.Illustrative image: a Mac mini (M4, 2024) on a store table.
$1,500 and up: big local models
Mac mini M6, 32 GB$1,799Cheapest Apple box that runs 27-35B local models alongside the agent. Apple M6, 12-core CPU, 12-core GPU, 170 GB/s, 32 GB memory, 1 TB SSD.Illustrative image: a Mac mini (2024 design).
Mac Studio M5 Max, 36 GB$2,499Faster local 27-35B models than a Mac mini (about 2.7x the memory bandwidth), still near-silent. Apple M5 Max, 18-core CPU, 32-core GPU, 460 GB/s, 36 GB memory, 512 GB SSD.Illustrative image: a Mac Studio (2022 design).
Mac mini M5 Pro, 64 GB$3,199Apple box that runs 70B-class local models at 4-bit. Apple M5 Pro, 18-core CPU, 20-core GPU, 64 GB memory, 1 TB SSD.Illustrative image: the back of a Mac mini (M4, 2024).
Illustrative image: a drawing of a cube-shaped mini PC.
Illustrative image: a drawing of a cube-shaped mini PC.
Amazon links on this site are affiliate links: if you buy, I may earn a commission, at no cost to you. As an Amazon Associate I earn from qualifying purchases. Nobody pays to be picked. Prices are the maker's list price, or a typical price, on 18 October 2026.
Self-hosted means the agent runs on a computer you control: its memory, its files and the accounts it signs in to live there. It doesn't usually mean the AI is there too. Most people connect the agent to a model in the cloud, which still sees what the agent sends it. Only a local model keeps that at home, and it takes far more memory.

I run my own agents on five computers: two old mini PCs, two office-surplus Dell OptiPlex desktops and a newer mini PC. Pictured: a Lenovo ThinkCentre Tiny beside a coffee mug, for scale.
Running costs
Leaving a Mac mini on all month costs about $0.53 in electricity, an old laptop about $0.86. Claude Code needs 4 GB of memory. A 30B local model needs 32 GB.
Idle power from published measurements, at the US average of 18.31¢ a kWh (EIA, July 2026). Memory from the Claude Code and Ollama docs. All the figures.
Longer answers
- Do you need a Mac mini for OpenClaw?
When a Mac mini is worth it, which one, and when the computer you have is enough.
- How much OpenClaw costs
The model, the computer and the electricity, worked out for light, daily and heavy use.
- Hermes Agent vs OpenClaw
The two most popular self-hosted agents, compared on setup, memory, channels and cost.
- OpenClaw alternatives
Hermes, NanoClaw, ZeroClaw, PicoClaw, nanobot and the coding agents, and the computer each one needs.
- Is OpenClaw safe?
What can go wrong when an agent has your accounts, and the precautions that limit it.
- Which computer can run your agent?
Each agent on an old laptop, office PC, mini PC, Mac mini, Raspberry Pi or VPS, from its own documentation.
The figures
Computers you may already have
Any of these can run an agent whose model is in the cloud. They differ in what they cost to leave on.
| Computer | Power when idle | Running cost | Good to know |
|---|---|---|---|
| An old laptop | 6.4 W | $0.86 a month in electricity | Costs nothing extra if you already own it, but you must set it never to sleep, including with the lid closed, and it stops when it loses power or Wi-Fi. |
| An office-surplus desktop (e.g. Dell OptiPlex) | 13 W | $1.74 a month in electricity | Uses about three times the idle power of a Mac mini or N100 mini PC, though that is still under $2.50 a month at the US average electricity price. |
| A mini PC (e.g. Intel N100) | 4.8 W | $0.64 a month in electricity | Small and cheap to leave on, but with no dedicated graphics card a local model would run on the CPU, which Hermes' docs say works but responds more slowly. |
| A Mac mini (Apple silicon) | 4 W | $0.53 a month in electricity | Very low idle power, and it runs macOS, which OpenClaw's docs say you need for iMessage. |
| A Raspberry Pi 5 | 2.7 W | $0.36 a month in electricity | The lowest power draw here, but it is ARM-based, so some optional tools may lack builds, and it is too slow for local models. |
| A rented server (VPS) | – | $15–24 a month | Always on and reachable without touching your home network, but you pay every month and your agent and its files live on someone else's computer. |
| The computer I use every day | – | None extra | No extra cost, but the agent only works while that computer is on and awake, and it shares your files, accounts and screen. |
Which agents you can self-host
Each still needs an AI model: a paid plan or API account in the cloud, or a local model.
| Agent | Runs on | Memory |
|---|---|---|
| OpenClaw | macOS, Linux, Windows (native or WSL2) | 1 GB minimum, 2 GB recommended |
| Hermes Agent | macOS (Apple silicon), Windows 10/11, Linux, WSL2 | No published minimum |
| Claude Code | macOS 13.0+, Windows 10 1809+ or Windows Server 2019+, Ubuntu 20.04+, Debian 10+, Alpine Linux 3.19+ | 4 GB minimum |
| Codex | macOS, Linux, Windows | 4 GB minimum, 8 GB recommended |
Memory for a local model
The minimums the Ollama and Hermes Agent docs give for common model sizes, in billions of parameters (B). On an Apple silicon Mac the processor and graphics share one pool of memory, so the Mac's total memory limits the model size. On a PC, a model runs fast only if it fits in the graphics card's own memory.
| Model size | Memory needed | Download |
|---|---|---|
| 3–4B | 8 GB | about 2.5 GB |
| 7–8B | 8 GB | about 5.2 GB |
| 14B | 16 GB | about 9.3 GB |
| 30B+ | 32 GB | about 19 GB |
Sources
- OpenClaw docs: Install (system requirements)
- OpenClaw docs: Raspberry Pi (minimum and recommended hardware)
- OpenClaw docs: DigitalOcean ($6/month 1 GB Droplet)
- Hermes Agent docs: Platform support
- Hermes Agent docs: Installation
- Hermes Agent README ($5 VPS)
- Hermes Agent docs: Run Hermes locally with Ollama
- Hermes Agent docs: Run local LLMs on Mac
- Hermes Agent docs: Local models
- Claude Code docs: Advanced setup (system requirements)
- OpenAI: Codex CLI install
- openai/codex: Installing & building (system requirements)
- Ollama README v0.12.0 (RAM per model size)
- Ollama library: qwen3 (download sizes)
- Apple Newsroom: Apple unleashes M1 (unified memory)
- Apple: Mac mini tech specs
- Apple Support: Mac mini power consumption and thermal output
- Notebookcheck: Lenovo ThinkPad T480s (i5-8250U) review
- Notebookcheck: Minisforum UN100L (Intel N100) review
- ServeTheHome: Dell OptiPlex 7070 Micro review
- ServeTheHome: Dell OptiPlex 7090 Micro review
- ServeTheHome: Dell OptiPlex 7060 Micro overview
- Tom's Hardware: Raspberry Pi 5 review
- Jeff Geerling: New 2GB Pi 5 has 33% smaller die, 30% idle power savings
- EIA Electric Power Monthly Table 5.6.A (July 2026)
- DigitalOcean: Droplet pricing
- Vultr API: plans (vc2-2c-4gb)
- Hostinger: VPS hosting plans
- Electricity: US average residential price, July 2026 (EIA Electric Power Monthly, released 2026-09-24). Monthly cost = watts × 730 hours ÷ 1000 × price.
Sources for the machines and prices (37)
- Apple Store US: Buy Mac mini (per-config fullPrice + part numbers), checked 18 October 2026
- Apple: Mac mini tech specs (M6, M5 Pro; memory bandwidth), checked 18 October 2026
- Apple Support: Mac mini power consumption and thermal output (published 2026-09-21), checked 18 October 2026
- Apple Store US: Buy Mac Studio, checked 18 October 2026
- Apple: Mac Studio tech specs, checked 18 October 2026
- Apple Support: Mac Studio power consumption and thermal output (published 2026-09-21), checked 18 October 2026
- Beelink store: EQ14 Intel N150, checked 18 October 2026
- Beelink store: SER8 AMD Ryzen 7 8745HS, checked 18 October 2026
- Beelink store catalog (incl. 'OpenClaw Pre-Installed' SER9/GTR9/GTi models), checked 18 October 2026
- GMKtec store: G3S Intel N95, checked 18 October 2026
- GMKtec store: M6 Ultra Ryzen 5 7640HS, checked 18 October 2026
- GMKtec store: EVO-X2 Ryzen AI Max+ 395, checked 18 October 2026
- GMKtec store catalog (EVO-X3 $3,799.99; EVO-X5 Pro 192 GB $6,799), checked 18 October 2026
- Framework Desktop DIY Edition configurator (all Max 300 configs out of stock on 2026-10-18), checked 18 October 2026
- Amazon: Apple 2026 Mac mini M6 16GB/256GB (MHQK4LL/A), checked 18 October 2026
- Amazon: Apple 2026 Mac mini M6 24GB/512GB (MHQM4LL/A), checked 18 October 2026
- Amazon: Apple 2026 Mac Studio M5 Max 36GB/512GB (MHL64LL/A), checked 18 October 2026
- Amazon: Beelink EQ14 N150 16GB/500GB, checked 18 October 2026
- Amazon: GMKtec G3S N95 8GB/256GB, checked 18 October 2026
- Amazon Renewed: Dell OptiPlex 7070 Micro i5-9500T 16GB/256GB, checked 18 October 2026
- Amazon: Beelink SER8 Ryzen 7 8745HS 32GB/1TB, checked 18 October 2026
- Amazon: GMKtec M6 Ultra Ryzen 5 7640HS 32GB/1TB, checked 18 October 2026
- Amazon: GMKtec EVO-X2 Ryzen AI Max+ 395 128GB/2TB, checked 18 October 2026
- ServeTheHome: Dell OptiPlex 7070 Micro review (i5-9500T, 2020-09-20), checked 18 October 2026
- ServeTheHome: Beelink SER8 (Ryzen 7 8845HS) review, 2024-08-12, checked 18 October 2026
- ServeTheHome: GMKtec EVO-X2 review, 2025-07-11 (8-14 W idle; Llama 3.3 70B Q6_K at 147-160 W), checked 18 October 2026
- Notebookcheck (via Wayback): Framework Desktop vs GMK EVO-X2 idle/load (1-min idle, external monitor), checked 18 October 2026
- Phoronix (via Wayback): Framework Desktop power & thermals, 2025-08-18 (12 W wall idle, Linux), checked 18 October 2026
- NVIDIA: DGX Spark (64 GB coming soon via OEMs; 128 GB; model-size claims), checked 18 October 2026
- The PC Enthusiast: DGX Spark 64GB from $4,999, Oct 23 2026; 128GB FE $4,699 (2026-10-03), checked 18 October 2026
- Hermes Agent docs: Run local LLMs on Mac (9B sweet spot 8-16 GB; 27-35B need 32 GB+), checked 18 October 2026
- Hermes Agent docs: local Ollama setup (CPU speeds: 9B ~10 tok/s, 31B ~2-5 tok/s on CPU), checked 18 October 2026
- OpenClaw FAQ: providers and hosting (no Mac needed; local model 'usually no'; 2 GB+ RAM), checked 18 October 2026
- Ollama library: llama3.3 70B default 43 GB, checked 18 October 2026
- Ollama library: gpt-oss 20b 14 GB, 120b 65 GB, checked 18 October 2026
- Ollama library: qwen3 8b 5.2 GB, 14b 9.3 GB, 30b 19 GB, checked 18 October 2026
- Earlier site estimates: harness RAM minimums, local-model RAM table and idle-power class figures (not an independently verified external source), checked 29 September 2026
Questions
- Prices checked 18 October 2026.
- Nobody pays to be picked.
- Setup is booked and paid through Fiverr.
What is a self-hosted AI agent?
An AI agent that runs on a computer you control, such as your own PC, a Mac mini or a rented server, rather than in the agent maker's cloud. OpenClaw, Hermes Agent, Claude Code and Codex can all run this way. The AI model can still be in the cloud.
How much RAM does a self-hosted AI agent need?
With the model in the cloud, 4 GB is the minimum Claude Code and Codex publish, and Codex recommends 8 GB. OpenClaw's gateway alone runs in 1 GB. A local model needs more: about 8 GB for a 7–8B model and 32 GB for a 30B one.
Do I need a Mac mini to run OpenClaw?
No. OpenClaw runs on Linux, Windows and macOS. A Mac mini is worth it for iMessage, which needs macOS, or for running local models in its shared memory. Mac mini for OpenClaw goes through the choice.
Is a rented server (VPS) or a computer at home better?
A server is always on and doesn't touch your home network, but costs about $15–24 a month for 2 processors and 4–8 GB of memory, and your agent and its files live on someone else's computer. A spare computer at home costs under $2 a month in electricity to leave on.
Is it safe to run an AI agent on my own computer?
A separate computer and limited accounts reduce what a mistake or attack can reach; they do not make an agent safe. Anything it reads can try to steer it, and connected accounts still need protection. See Is OpenClaw safe? for risks and precautions.
Do you get paid for these picks?
Links to Amazon can earn me a commission if you buy; it doesn't change your price. Nobody pays to be listed, and Apple and Framework links go straight to the maker and earn nothing. I also sell setup, on Fiverr, which is labelled where it appears.
Why is there no Amazon price next to the button?
Amazon's prices change by the hour, so the button opens the live listing. The price beside each machine is the maker's list price on the date above, or a typical price where there is no list price.


- Tulsa, Oklahoma
- Runs agents on five computers
- About Tyler
I'm Tyler Mayberry. I build AI agents and the computers they run on. I run agents on five computers and write about choosing, setting up and maintaining them.
This site answers what people ask before they buy: which machine, what it costs to leave on, and how to keep an agent from doing damage. How it's made, and how it earns money.