A developer released an update to Pi Herdsman, an open-source extension for coordinating multiple coding agents. The new setup uses a Manager → Lead → Agent hierarchy where each agent runs in its own Git worktree with an independent context, keeping implementation history out of the orchestrator. The demo and source are public on GitHub.
A developer released Pi Herdsman, an open-source coordination layer that keeps a lead coding agent's context small while parallel workers handle investigation and implementation in separate Pi sessions. It separates conversation state (Pi), session ownership (Herdr), and orchestration, routing questions and results back through an ownership tree instead of accumulating every tool call in the lead's context.
A developer released Pi Herdsman, an open-source coordination layer that keeps a lead coding agent's context small while parallel workers handle investigation and implementation in separate Pi sessions. It separates conversation state (Pi), session ownership (Herdr), and orchestration, routing questions and results back through an ownership tree instead of accumulating every tool call in the lead's context.
Managing Agent Worktrees in Git: guide on cleanup and ephemeral lifecycles for parallel AI coding agents
Managing Agent Worktrees in Git: guide on cleanup and ephemeral lifecycles for parallel AI coding agents
A Reddit user describes the agent orchestration architecture that held up across real overnight coding runs. The design pairs one manager agent with multiple worker agents isolated in separate git worktrees, emphasizing strict role separation and one writer per worktree.
A Reddit user describes the agent orchestration architecture that held up across real overnight coding runs. The design pairs one manager agent with multiple worker agents isolated in separate git worktrees, emphasizing strict role separation and one writer per worktree.
A developer released Agenttik, a tool for managing multiple AI coding agents across parallel projects, announced via a Show HN post linking to GitHub. It lets users assign tasks to agents from different providers (Claude Code, OpenAI subscriptions), queue and schedule work, and review changes in-tool. The project was built in one week after the author found parallel agent workflows mentally exhausting without organizational support.
A developer released Agenttik, a tool for managing multiple AI coding agents across parallel projects, announced via a Show HN post linking to GitHub. It lets users assign tasks to agents from different providers (Claude Code, OpenAI subscriptions), queue and schedule work, and review changes in-tool. The project was built in one week after the author found parallel agent workflows mentally exhausting without organizational support.
A solo developer describes AIPass, an open-source infrastructure for persistent AI agents that plan, execute, review, and maintain codebases without manual coordination. A lead agent named DevPulse dispatches tasks to specialists like @hooks and @memory, each with persistent identity, memory, and mailboxes. The system is being integrated with Codex alongside Claude Code so agents can run across model providers.
A solo developer describes AIPass, an open-source infrastructure for persistent AI agents that plan, execute, review, and maintain codebases without manual coordination. A lead agent named DevPulse dispatches tasks to specialists like @hooks and @memory, each with persistent identity, memory, and mailboxes. The system is being integrated with Codex alongside Claude Code so agents can run across model providers.