Decide what to work on next

You finished something. Instead of reading issue summaries, or scouring sticky notes and text files, ask kcap what is waiting on you, what you left unfinished, and what is new, with the reason for each. Ask from any coding agent kcap supports; the example uses Copilot CLI

Ask Copilot CLI, see the ranked list with its reasons, snooze one in the web UI, hand one to an agent, and leave a loose end for next time. Everything below stands on its own

In your agent

What should I work on next?

The agent answers from a ranked list of things that need your attention: loose ends from your previous work, sessions you interrupted, work a teammate is waiting on, review requests, high-priority issues. Each row says why it is there

Pick one:

Take the second one. Claim the loose end and continue from where it stopped

Or decline:

Dismiss the first; that migration was abandoned

A dismissed row stays dismissed until you restore it

In the web app

The Next work band at the top of the dashboard shows the same feed. Each row has the reason and the evidence behind it: the session, the work item, who is waiting. From a row you can:

• Snooze a loose end until tomorrow, three days or next week
• Dismiss a suggestion, and later Restore it from the dismissed table
• Work on this to hand the row to a hosted agent: pick the daemon and checkout, the harness and model, and the prompt the agent will receive

Every one of these is also a sentence to your coding agent: “snooze that until next week”, “dismiss the first”, “start an agent on the second one”

Leaving loose ends on purpose

The feed is only as good as what agents put in it. When you stop a session short, say so:

I'm stopping here. Leave a loose end for the retry test that still needs writing

The agent records a concrete, claimable piece of work that outlives the session and the chat

Further reading

• Next work and loose ends
• Work items