Lume
Open Lume
Claude Code · OpenAI Codex

Assign a to-do. Let the agent do it.

Hand a task straight from your list to Claude Code or OpenAI Codex. It works in the background, narrates progress into the task, and returns the result for your review. You still hold the checkbox.

Open Lume free →Read the Mac mini loop guide
Assigned
You hand “Refactor billing webhook” to Claude Code.
Doing
The agent reads the repo and works — progress streams into the task.
Ready for review
It returns a result and a link. Nothing is marked done.
You check it off
You review, accept, and tick the box — or send it back.
How it works

A task queue your agents actually understand

Lume exposes your tasks over MCP and a REST API, so a coding agent can pick up what you assign it, do the work, and report back — no glue code required.

Dispatch from the list

Assign any to-do to Claude Code or Codex the same way you’d assign it to a teammate. The agent gets the title, notes, and context.

Progress in the timeline

The agent posts updates into the task as it works, so you can see what it’s doing without leaving your list.

Human-in-the-loop by default

Agents mark work “ready for review”, never “done”. You decide what ships — with full undo if you change your mind.

Run it on a loop

Point a Mac mini at your list and walk away

Because Lume has a real API and MCP server, you can run Claude Code or Codex in a loop — polling for anything you’ve assigned them, doing it, and handing it back. Your list clears overnight.

How I built mine →
while true; do
  task=$(lume next --assignee codex)
  codex run "$task"
  lume update "$task" --review
done

Assigning tasks to AI agents — FAQ

Can I assign tasks to Claude Code?+

Yes. In Lume you assign a to-do to Claude Code just like assigning it to a person. The agent picks it up over MCP/API, works on it in the background, and returns the result tagged “ready for review”.

Does it work with OpenAI Codex too?+

Yes. Codex is supported the same way as Claude Code — assign a task and the agent does the work, posting progress back into the task’s timeline.

Will an agent ever mark my task done?+

No. Agents only ever set a task to “ready for review”. You stay in control of what’s actually finished, and every change is undoable.

How do I run agents automatically on a schedule?+

Use Lume’s API or MCP server to poll for assigned tasks from a machine like a Mac mini, run Claude Code or Codex on each, then mark them for review. Our blog has a full walkthrough.

Delegate the busywork to your agents.

Free to start. Assign your first task to Claude Code or Codex today.

Open Lume free →