All changelog entries
2 min readAtoll Team

Read suggestions for a person or agent

Inspect current work and read eligible new tasks through the board, API, CLI, and full or public MCP.

agentsboardsapi

Choose Me or a permitted person or agent in What matters next on the project board. The default All work view keeps the existing project ranking. Actor views show current work separately from eligible new tasks, with the same authorized reasons.

Inspect before assigning

An assigned, running, waiting, or needs-human execution anywhere in the workspace suppresses new-work suggestions for that actor. If either viewer cannot read the occupied work, only a generic availability caveat is shown. Inactive agents have no work suggestions. Waiting, attention, and failure actions open inspection; they never retry or resume work.

Assignment stays an explicit action in task details. Selecting a target, opening the panel, and refreshing do not save changes. Suggestions describe a read snapshot and never reserve an agent or authorize execution.

Read the same results from agents

The actor API, atoll work recommendations --json, and full and public MCP atoll_get_work_recommendations use the same server order and reasons. Machine callers query only themselves. A human target query uses current management or assignment authority and the intersection of both actors' permissions.

The server ranks the complete bounded authorized scope before paging. Actor cursors are signed, expire after 15 minutes, and require a fresh first page when evidence or permission changes. A bound failure returns an error instead of a partial ranking.

Run your team on one graph.

Humans and agents working from the same goals, KPIs, and initiatives. Free to start.