Show HN: Moadim.io – A scheduler for agents
Moadim.io introduces an open-source Rust daemon designed to simplify the scheduling and management of AI agent routines locally. It leverages Git for version-controlled "loop engineering," allowing users to define prompts, schedules, and agents in an isolated, reboot-resistant environment. Hacker News finds value in its practical approach to agent orchestration, offering a local, agent-agnostic alternative to cloud-based solutions, particularly appealing to those grappling with agent workflow automation.
The Lowdown
Moadim.io is presented as a novel solution for orchestrating AI agents, focusing on local, open-source, and Git-managed routine execution. This "loop engine" aims to provide a reliable framework for automating tasks performed by various AI models, addressing the need for robust agent scheduling outside of complex cloud infrastructures.
- Core Functionality: Moadim is a local Rust daemon that acts as an agent scheduler, allowing users to define and manage AI agent routines through a Git repository.
- Key Features: It's 100% open source, agent-agnostic (supporting Claude, Codex, Hermes, Pi, etc.), OS-agnostic (macOS, Linux with tmux), and offers multi-runner support.
- "Loop Engineering": The project introduces the concept of "loop engineering," where a loop combines a prompt, a schedule, and an agent, executed in an isolated workbench with a watchdog for reliability.
- Deployment & Management: Installs as a service (launchd/systemd) to survive reboots, uses tmux for agent execution isolation, and provides UI, REST, and MCP interfaces for interaction.
- Philosophy: The author emphasizes keeping the software "done" and laser-focused on agent scheduling, avoiding feature creep into areas like webhooks.
By offering a self-hosted, version-controlled, and agent-agnostic system for automating AI agent tasks, Moadim.io positions itself as a foundational tool for developers looking to build more sophisticated and reliable agent-driven workflows without external dependencies.
The Gossip
Cron-like Choreography for Code Agents
The primary discussion revolved around understanding Moadim's core utility, with users likening it to a "userland cron/launchd" for agents. The author affirmed this, explaining that Moadim simplifies managing agent routines in a version-controlled Git repository, handling agent spawning, worktree setup, and process states without manual configuration of lower-level cron jobs. This resonated with users who had struggled with similar problems, like versioning `plist` files for agent access to `launchd`.
Scoping the Scheduling Solution
Commenters probed the boundaries of Moadim's functionality, particularly concerning token budget management for agents and integration with event-driven actions like webhooks or LLM observability. The author clarified that Moadim intentionally remains agent-agnostic and focuses exclusively on the scheduling aspect, leaving token economics to the agents themselves and purposefully not branching into webhooks. However, the author expressed openness to suggestions for observability features, acknowledging it as a pain point.
Practicality and Proactive Participation
The author garnered positive attention by demonstrating Moadim's immediate utility in monitoring the Hacker News discussion itself. They disclosed having a Moadim routine that polled the thread every 10 minutes and sent Discord notifications for unaddressed comments, showcasing a direct, practical application of the scheduler for real-time engagement.