Plan in the open
The agent's todo list as it thinks: what's done, what's running, what failed and why. Sub-tasks, live timers, a folded history.
<TaskList />Runs, plans, terminals, diffs, traces and approvals. 15 open-source React components for supervising coding and computer-use agents, with a manifest your own agent can install from.
The agent's todo list as it thinks: what's done, what's running, what failed and why. Sub-tasks, live timers, a folded history.
<TaskList />Accept or reject the agent's edits hunk by hunk, with syntax highlighting, word-level changes and j/k/a/r on the keyboard. Parses real unified diffs.
<DiffView />Risky commands wait for a human. Risk level, the exact command, details on demand, ⌘↵ to approve and Esc to deny.
<ApprovalPrompt />Clears stale output before rebuilding. Affects 1,204 files under ./build only.
rm -rf buildEvery model call, tool and subagent on one time axis. Running spans grow live; select one for its input, output, tokens and cost.
<TraceWaterfall />Computer-use made legible: the page, the agent's cursor and target, and a step log. Take over when it needs you.
<BrowserSession />Create a new pull request by comparing changes across two branches.
4 commits · 6 files changed · +142 −38
Queue follow-ups while the agent works. Reorder, edit, or send one now to interrupt and redirect the run.
<PromptQueue />Alt plus arrow keys or Space then arrows to reorder. Enter to edit, Delete to remove, Control or Command plus Enter to send now.
Hover a prompt, or focus it and use Alt+↑/↓, Enter, Delete and ⌘↵.
Every component ships a typed manifest: when to use it, when not to, every prop, the anatomy and the keyboard map. The registry, llms.txt, the MCP server and the CLI are generated from it.
# give your coding agent the catalog
claude mcp add kura -- npx -y @kura-ui/mcp
# then ask for what you need
> add per-hunk review for the agent's edits, and a terminal for its test runs
# or install directly
npx shadcn@latest add https://kuro.abstergo.fyi/r/diff-view.json