AI Agent Task Manager - Run multiple AI coding agents
The AI Agent Task Manager is a dedicated task board designed for running multiple AI coding agents in parallel across separate workstreams. Each column on the board maps to an autonomous agent session such as Claude Code, Codex, or Cursor, allowing developers to organize queues of tasks with detailed instructions. Agents poll the board via a simple REST protocol, pull their next assigned task, execute it autonomously, and report back when finished or blocked.
Key Features
- Workstream Columns: Separates parallel agent execution into independent columns, with each stream representing a distinct project or terminal session.
- Task Instructions Queue: Allows users to queue individual cards containing full instruction bodies that agents receive verbatim upon picking up a task.
- Agent REST Protocol: Exposes standardized endpoints for agent interaction, supporting task polling, attention flag raising, resumption, and completion reporting.
- Attention Notifications: Highlights columns amber and triggers OS-level web push notifications whenever an agent encounters a blocker requiring human input.
- Long-Poll Integration: Minimizes unnecessary polling traffic by keeping connections parked up to 30 minutes until a new task enters the active queue.
- Cross-Device Sync: Persists boards across refreshes and devices through a signed-in account model while retaining an anonymous browser-stored mode.
Built for developers, indie hackers, and technical founders juggling multiple autonomous coding workflows simultaneously, this utility removes the need to constantly hunt across active terminals. It provides a structured management layer for automated software development tasks without locking users into specific proprietary agent platforms. Available as a free tool with no signup required for anonymous browser boards and optional sign-in for cross-device syncing.
