Topics / #on-premise

On-premise projects on GitHub

Self-hosted software designed to run on private infrastructure instead of cloud.

As of , the most-starred On-premise projects on GitHub are private-gpt (57.6K stars), gitlabhq (24.6K stars) and openmed (5.7K stars). GitHub.Locker tracks 5 On-premise repositories with 93.8K stars combined. The fastest-growing on October 16 was private-gpt, with +7 stars in a day.

5Projects
93.8KStars in total
+13Stars in a day

Hot right now

Most stars gained in a day · Oct 16

The all-time top

5 projects · by stars

  1. #1 · zylon-ai/private-gpt

    private-gpt

    PrivateGPT is an open-source API layer that turns any local, OpenAI-compatible model into a full production-ready AI application backend, covering RAG, tools, MCP, text-to-sql and more.

    57.6K
  2. #2 · gitlabhq/gitlabhq

    gitlabhq

    This is GitLab Community Edition, an open-source platform for managing Git repositories with merge requests, issue tracking, wikis, and built-in CI/CD pipelines, used by more than 100,000 organizations.

    24.6K
  3. #3 · maziyarpanahi/openmed

    openmed

    OpenMed is a local-first healthcare AI SDK: clinical NER and HIPAA PII de-identification that runs on your own hardware, with 2,200+ medical models and Python, MLX, Android, and browser runtimes.

    5.7K
  1. 4

    astron-rpa

    5.3Kiflytek/astron-rpa · Python

    AstronRPA is an open-source, enterprise-grade Robotic Process Automation (RPA) desktop application with a visual designer that lets you build low-code/no-code workflows to automate desktop software and web pages, with native integration into iFlytek's Astron Agent platform.

  2. 5

    agent-safe-pipeline

    688decionis/agent-safe-pipeline · TypeScript

    Agent-Safe Pipeline is a reference implementation of an 'Execution Authority' architecture that sits between AI agents or APIs and consequential HTTP actions: it intercepts the action, asks Decionis whether it is authorized, and only forwards it if a single-use grant is issued, otherwise holding it for a human or refusing it, with every decision leaving tamper-evident evidence.

Often tagged together