GitLearnOS

Native DeepSeek Harness support · A GitLearnOS-exclusive learning panel

Quickstart: give one request to your AI →

中文 · Website · Documentation map · Protocol

GitLearnOS gives an AI a learner-owned Git memory: notice real learning, organize useful evidence, guide the next step, and write results back automatically.

Learning may happen with teachers, class, paper, books, practice platforms, projects, peers, or AI. GitLearnOS does not move all learning into one app. One replaceable main agent connects the useful evidence and next actions inside the learner's own Git repository.

Git stays in the background. The learner does not need to manage folders, commits, branches, or a Git hosting service during normal learning.

It should notice learning without a command

Once configured, GitLearnOS should not wait for “use GitLearnOS,” “save this,” or an explicit Skill invocation. A subject question, attempted answer, photographed page, class note, teacher comment, practice result, or repeated difficulty can be a learning event.

The agent answers the immediate need first. Under safe-auto, it then makes the smallest useful writeback when value, target, and privacy are clear. If the event is ambiguous it makes one brief suggestion or asks one necessary question. Incidental conversation is not stored.

Cross-conversation continuity comes from:

flowchart TD
    A["Project rules"] --> C["Proactive agent"]
    B["Native memory"] --> C
    D["Project Sources"] --> C
    E["Current chat"] --> C
    C --> F["Learner Git state"]

Skills improve a workflow but are never the only activation mechanism. The setup installs the compact project/custom instructions and, when permitted, a verifiable , so the core loop survives on a surface that does not expose Skills.