Back to directory

MindIE Agent

A collection of plugins and a shared knowledge runtime that give coding agents such as Codex, Kimi Code and Claude Code domain context, stable remote tools and a feedback loop of accumulated experience for vLLM and vLLM-Ascend development tasks.

MindIE Agent

MindIE Agent is split across several repositories: this one holds the architecture, domain boundaries and delivery entry point, while separate repositories provide native plugins, hooks and task integration for Codex, Kimi Code and Claude Code, a shared knowledge and experience runtime, domain content for vLLM-Ascend, a remote execution environment, and diagnostics.

Its first supported domain is vLLM and vLLM-Ascend development. Knowledge collection, and once contribution is enabled its anonymization, both run locally; remote execution is provided through a separate remote-development component. The project describes itself as being in a pre-release restructuring, with data collection off by default.

Each of the three agent adapters has reached native use on macOS, with evidence of cross-version task handoff and rollback after a failed install; Windows support is being advanced separately. Older workspace, bootstrap and auto-update components have been removed from the current tree, with no aliasing for old names.

More in Agent skills

See all 581
  • Provenance gate

    github.com

    A deterministic authorization gateway that sits between an AI agent and its tools, tracing where each parameter of a side-effect call came from, and blocking the call or asking a human when the source does not check out.

  • A markdown framework of system rules that restates Isaac Asimov's Four Laws of Robotics as a set of prioritized loss calculations for large-language-model agents to follow, published as a discovery file for agents and crawlers to read.

  • Vitruvius

    github.com

    An engineering research agent that runs a discover, read, synthesize, verify and review loop over technical questions, reading sources directly rather than inferring from titles, and publishing its own scored verifier benchmark, weaknesses included.