What the box does, and what it doesn't
Get this split clear before you buy anything. When you run Claude Code or an app built on the Claude Agent SDK, the model runs in Anthropic's cloud and you reach it over the API. Claude Code's own requirements list an internet connection as mandatory, along with a Pro, Max, Team, Enterprise or Console account. The mini PC runs everything around the model. That's the agent loop, the shell it runs commands in, git, your test suites, MCP servers, headless browsers, and the scheduler that kicks jobs off at 3 a.m. So a modest used box does the job, and you don't need a GPU for it. Running the model itself on your own hardware is a different and much heavier workload, covered in running local LLMs on a mini PC.
Claude Code and Agent SDK requirements
Claude Code's published system requirements are modest. It runs on macOS 13+, Windows 10 1809+ or Windows Server 2019+, Ubuntu 20.04+, Debian 10+ or Alpine 3.19+, with 4 GB+ RAM and an x64 or ARM64 processor. The native installer doesn't need Node.js at all. Only the optional npm package asks for Node.js 22 or later. If you build your own agent with the Agent SDK, the runtime is Python 3.10+ or Node.js 18+, and the SDK bundles its own Claude Code binary.
The SDK docs earn their keep on sizing. Each agent session is a separate long-lived subprocess with its own shell, working directory and transcript on disk. Anthropic suggests 1 GiB RAM, 5 GiB disk and 1 CPU per agent as a starting point, and says memory grows with session length and tool activity. It also gives a simple formula, where agents per host equals host RAM minus overhead, divided by the peak RAM one session actually reaches. Measure a real session with your real tools instead of trusting the floor. In practice, 32 GB is a comfortable target for a few agents at once, and 64 GB is where browser-heavy fleets stop fighting each other.
Scheduled agents with cron or systemd timers
For unattended jobs, start with Claude Code's non-interactive mode. `claude -p "…"` runs one task and exits with code 0 on success and non-zero on failure, so cron or a systemd timer can branch on the result. The docs recommend `--bare` for scripted runs. It skips auto-discovered hooks, plugins and MCP servers, and it needs `ANTHROPIC_API_KEY` set. `--output-format json` returns the result plus an estimated cost figure, and `--permission-prompts none` denies anything that would otherwise wait for a human. Add a tight `--allowedTools` list and a dedicated Linux user, and a nightly dependency-update or report-writing agent takes a few lines of config.
Browser agents are the heavy part
Agents that drive a real browser through Playwright are the hungriest thing you'll put on this box, because every session is a full Chromium instance. Playwright currently requires Node.js 22, 24 or 26, and on Linux it lists Debian 12/13 and Ubuntu 22.04, 24.04 and 26.04 on x86-64 or arm64. Playwright doesn't publish a RAM figure per browser, so measure. Launch your usual number of sessions, watch peak memory, and size up from there. That's the main reason the picks on this page lean toward 64 GB-capable machines. Ubuntu 24.04 LTS is on both Claude Code's and Playwright's support lists, which makes it the easy OS choice.
MCP servers and webhooks
MCP servers come in two standard transports. With stdio, the agent launches the server as a local subprocess, which costs a little RAM per server and nothing else. With Streamable HTTP, the server is an independent process with its own endpoint. The spec says local servers should bind to 127.0.0.1 instead of every interface and should authenticate connections. Keep them on localhost and you avoid most of the risk.
Webhooks, like a GitHub push, a form submission or a payment event, need a way in from the internet. Use a tunnel instead of forwarding a port on your router. Cloudflare Tunnel's `cloudflared` makes an outbound-only connection, so the box needs no publicly routable IP and you can block all inbound traffic. Tailscale Funnel routes public traffic to a local service on ports 443, 8443 or 10000, TLS only. It's available on all Tailscale plans and has bandwidth limits you can't configure, which is fine for webhooks.
Why a mini PC beats a laptop or a VPS for this
- An old laptop wants to sleep when the lid closes, rides on Wi-Fi, and keeps its battery at full charge for months. A mini PC has none of those habits, takes a wired connection, and can be set in BIOS to power back on after an outage.
- A cloud VPS bills you every month, forever. A mini PC is a one-time purchase plus electricity, and your repos, browser profiles, API keys and the files agents work on stay in your house instead of on someone else's disk.
- A VPS still wins on redundant power, a static IP and nobody tripping over the cable, because it lives in a data center. If your agents serve paying customers, host the public endpoint in the cloud and let the home box do the private work.
Power is the running cost that matters. At the 11 to 14 W a ThinkCentre M720q draws at idle, the yearly electricity estimate on this page stays small, and agent bursts barely move the average. Anthropic's own hosting guide points out that model token spend typically dwarfs infrastructure cost, so the box will rarely be your biggest line item.
How much machine you need
- One or two scheduled agents with no browser run fine on any 8th-gen 1-liter tiny with 16 GB, such as the ThinkCentre M720q.
- Several agents plus Playwright want six or more cores and 32 GB.
- For a small fleet of browser agents, or agents next to Proxmox VMs and workflow tools like n8n, get a 64 GB-capable box such as the ThinkCentre M75q Gen 2.
Choosing between the three big business brands? Our ThinkCentre vs OptiPlex vs EliteDesk comparison covers how they differ on RAM limits, storage layout and remote management.






























