Heroboard, Part 1 — development management for people and AI agents

Task

AI agents can already explore code, implement features and run tests, yet project management often sits outside that loop. Developers change code with an agent and then manually copy the same context into cards, comments and statuses. Tasks become stale and decision history is lost between the terminal and the board.

Project story

1. Every project gained its own process

   Each project stores its workflow and context: rules, environment, validation and release order. Kanban reflects the real path from backlog to release.

2. The AI agent joined the workflow

   Through MCP, the agent reads project context, finds or creates a task, starts work, reports progress and moves the result to the next stage.

3. A task no longer ends at “code complete”

   Review, Deploying, QA and Released are part of one history. Releases connect completed tasks with delivered value.

4. Context survives a restart

   Agent work sessions link back to tasks. After a restart, the working structure can be recovered and continued from the current state.

What we built

• →Projects, epics, tasks and configurable Kanban columns
• →Drag-and-drop and manual management in the web interface
• →MCP and Agent API for AI tools
• →Agent-driven task creation, status changes and progress reports
• →Stored project process and technical context
• →Agent sessions linked to tasks for recovery
• →Roadmap, planned dates and releases linked to completed tasks
• →Team roles and a separate client portal

The product in use

Result

Heroboard became a shared work surface for developers, teams and AI agents. Project management happens in the same loop as implementation: a task exists before work begins, progress is captured along the way, and the result reaches review, QA and release. People no longer have to act as secretaries for their own AI, while control points remain visible.

Stack

Related services