Open-weight models are now good enough that keeping inference in-house is a practical choice rather than a compromise. These projects cover local model runners, private assistants and the self-hosted infrastructure around them.
The argument for running AI locally used to be ideological and is now mostly commercial: no per-token bill, no data leaving the network, and no vendor deprecating the model your product depends on. Regulated industries reached that conclusion first; everyone else is arriving via the invoice.
What the category still lacks is the boring layer — updates, monitoring, access control, backups. The runners work. Turning a runner into something a team can depend on is where the open problems are, and several projects below are attempts at exactly that.
23 analysed projects match this topic.
Skip weeks of custom work by starting from a working artificial intelligence foundation.
Skip weeks of custom work by starting from a working screen recording foundation.
Skip weeks of custom work by starting from a working image generation foundation.
Skip weeks of custom work by starting from a working ai agent foundation.
Skip weeks of custom work by starting from a working social media management foundation.
Skip weeks of custom work by starting from a working artificial intelligence foundation.
Skip weeks of custom work by starting from a working email service foundation.
Skip weeks of custom work by starting from a working ai agent foundation.
Skip weeks of custom work by starting from a working whoop foundation.
Skip weeks of custom work by starting from a working windows optimization foundation.
Skip weeks of custom work by starting from a working map tracker foundation.
Skip weeks of custom work by starting from a working wardrobe organizer foundation.
Skip weeks of custom work by starting from a working podcast foundation.
Skip weeks of custom work by starting from a working php foundation.
Skip weeks of custom work by starting from a working bookmarking foundation.
Skip weeks of custom work by starting from a working docker compose foundation.
Skip weeks of custom work by starting from a working infrastructure foundation.
Skip weeks of custom work by starting from a working operating system foundation.
Skip weeks of custom work by starting from a working ai agent foundation.
Skip weeks of custom work by starting from a working inbox scanner foundation.
Skip weeks of custom work by starting from a working price tracker foundation.
Skip weeks of custom work by starting from a working ai agent foundation.
Skip weeks of custom work by starting from a working llm foundation.
Updated: 2026-08-17