Awesome Infra for AI › Workflow Orchestration for AI

holon-run/holon

⭐ 150 Rust repository created 2025-12-18

Holon is a local-first workbench designed to provide a continuous working environment for AI agents. Unlike typical agent executions that might restart with each session or command, Holon ensures that agents maintain their state, context, and progress across sessions, commands, human input, and external events. It is not an agent itself but rather an infrastructure that supports multiple agents, enabling them to tackle complex, long-running tasks. Key features include a continuous agent workspace, which allows agents to preserve their working context; a "work-first" task model that organizes tasks, waits, and execution progress; and event-driven wait and wake mechanisms, where agents can pause and resume based on task results, external triggers, or operator input. Holon also emphasizes explicit context and trust boundaries, ensuring that input from various sources (operator, tools, external events) is clearly distinguished. It is built for seamless integration with local development environments, supporting local repositories, shells, and toolchains. The project handles provider setup, supports various LLM providers (Anthropic, OpenAI, DeepSeek, etc.), and offers both a command-line interface (CLI) and a Terminal User Interface (TUI) for interaction. Core concepts include Agents (long-lived identities), WorkItems (advanceable goals), Tasks (asynchronous execution), and the ability to define wait conditions. Holon's primary purpose is to keep agent work alive in a local workspace, making it observable, resumable, waitable, delegable, and deliverable, fostering robust and persistent AI agent applications.

https://github.com/holon-run/holon

agent-runtimeai-agentsai-infrastructuredeveloper-toolslocal-firstworkflow-orchestrationllm-ops