2026.

Raspberry Pi 5, mini PC, Mac mini or Mac Studio for a local LLM or a self-hosted AI agent: what each budget buys, with makers' specs and published speeds.

PocketClaw's Q2 2026 look at self-hosted AI: where the serious CVEs landed, with a link to NVD for each, and our editorial view on agents, hardware, providers, models and infrastructure. The star counts, reader survey and market shares of the first version have been removed: they had no source.

Self-hosted AI agent at the centre of the January 2026 security crisis. Its tool sandbox is off by default: the main session's tools run on the host until you configure one.

Personal AI agent in Python (MIT) with a browser UI and chat apps. Native Anthropic, OpenAI and OpenRouter providers plus local servers; its shell sandbox is off until you switch it on.

Small TypeScript agent (MIT) built on Anthropic's Claude Agent SDK. Every agent runs in its own Docker container, on macOS, Linux or Windows through WSL2.

Rust agent from NEAR AI, MIT or Apache-2.0. Untrusted tools run in a WebAssembly sandbox with declared capabilities, and secrets are injected outside them.

Single Rust binary from ZeroClaw Labs, MIT or Apache-2.0. Works with Anthropic, OpenAI, Ollama and about twenty other providers, and turns on an OS sandbox it detects.

Cloudflare's experimental package that runs OpenClaw inside a Cloudflare Sandbox container behind a Worker. Needs the Workers Paid plan.

Managed assistant built on OpenClaw, in Telegram and Discord: $20 a month self-serve, or $100 with setup and a monthly call. AI credits included.

Managed OpenClaw hosting, live on Telegram (Discord and WhatsApp "Soon"). $19 or $49 a month, $14 or $37 billed yearly. Your own model API key.

Raspberry Pi 5 for self-hosted AI: four Cortex-A76 cores, 1 to 16 GB of LPDDR4X. Hosts Hermes Agent (no browser tool) or Nanobot; local models up to about 3B.

Mac mini M4 and M4 Pro for local LLMs: 16 to 64 GB of unified memory at 120 or 273 GB/s, the two specs that decide which models load and how fast.

€20 SBC with 512 MB RAM. Useful as an edge agent endpoint (sensor reader, tool host) but not a primary agent host.

Lenovo ThinkCentre M75q, used: a €150-250 Ryzen mini PC with 16 GB of RAM and an NVMe slot, enough for Hermes Agent with the browser tool.

€100-180 used Intel mini PC. 8th gen i5, 8-16 GB RAM, NVMe. The cheapest credible mini PC for self-hosted AI.