ram.

Configure V8 heap limit to prevent OOM kills on 1 GB Android device, starting at 384 MB and refined down to 112 MB.

Replace unused ESM channel SDKs with stub packages exporting empty classes, saving ~80 MB RAM and 262 MB disk.

Systematically binary-search the minimum V8 heap cap that survives OpenClaw startup on 1 GB RAM.

Defer 40 MB of module loading at startup using ES6 Proxy-based lazy require with 37 intercepted package prefixes.

Splitting intercepted packages into dead (instant empty stub, zero RAM) and lazy (proxy-deferred, loaded on demand), with 23 packages stubbed out completely to save memory and I/O on a 1 GB phone.

Establish 192 MB as the minimum V8 heap for OpenClaw startup, surviving the ESM module linking spike.

Shrink V8 young generation semi-space from default to 1 MB, saving RAM on I/O-bound gateway with minimal GC overhead.

Reduce libuv thread pool from 4 to 1 on single-core ARM32, saving 2 MB RAM with zero performance impact.

Remove 31 stock Android packages without root using pm uninstall, freeing ~100 MB RAM on Moto E2. Permanent on Android 6.

Configure static IP to decouple WiFi from Google Play Services, enabling safe GMS kill to reclaim 270 MB.

Second debloat pass adding Chrome, defcontainer, and qcrilmsgtunnel to the disabled list, saving additional 100 MB RAM.

Replace WebView launcher with native Android Activity using ScrollView and TextView, cutting RSS from 216 MB to 55 MB.

Redesigned dashboard with green CRT theme and per-process RAM breakdown that revealed WebView as the #1 memory consumer.

Merge three separate background bash kill loops into a single loop, saving 3 MB RAM and reducing process count.

Cron job to periodically kill com.termux.boot Dalvik VM, freeing 40 MB while avoiding the cgroup cascade from killing com.termux.

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.

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.

Managed OpenClaw hosting for Telegram, Discord and WhatsApp. No free plan: $20, $49 or $200 a month, with model usage bought separately as credits.

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.

Intel NUC 13 and Geekom IT13: 13th-gen Intel Core mini PCs with 16 to 64 GB of RAM, room for Hermes Agent with browser automation and a 7B model beside it.

Repairable, modular laptop with strong Linux support. The right choice if you want a portable agent host that doubles as a daily driver.

The PocketClaw origin story device. €15-30 used. Runs lightweight agents via Termux + proot, with serious caveats.

Orange Pi 5 Plus: Rockchip RK3588, 4, 8 or 16 GB of LPDDR4/4X and a 6 TOPS NPU. The Raspberry Pi 5's closest competitor for self-hosted AI on ARM.

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.

The Khadas Edge2 in its Pro configuration: Rockchip RK3588S2, 16 GB of LPDDR4X and 64 GB of eMMC on a small board. Polished hardware, smaller community.

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