Open-source React components for AI agent conversations, reasoning, tools, approvals, and streamed responses.
One adaptive activity stream for reasoning, searches, tool calls, structured execution traces, or a chronological mix.
A compact activity glyph, action verb, and live tabular timer for longer-running agent work.
Shimmering reasoning copy with an ASCII loader and cascade, phrase-swap, or per-letter scramble transitions.
A quiet shimmer that keeps the agent's current status readable while work continues.
A collapsible AI workspace sidebar for folders, projects, files, and bookmarks with keyboard navigation, optimistic moves, inline rename, and overflow-aware labels.
A human-in-the-loop decision surface for approvals, single or multiple-choice questions, custom responses, and multi-step review flows.
A complete agent conversation workspace composing navigation, messages, streaming, planning, approvals, tools, code, diffs, generated media, sources, and prompt input.
Inline citation markers paired with a collapsible, progressively rendered reference collection for grounded agent responses.
A syntax-highlighted code surface with stable streaming updates, line numbers, focused lines, smooth following, and copy feedback.
A syntax-highlighted file change disclosure with progressive rows, line numbers, live change counts, smooth following, and completion collapse.
A stable generated-image surface that moves from queued work through progressive refinement to a completed result without layout shift.
Composable conversation primitives for message rows, grouped bubbles, avatars, metadata, live markers, and a mount-only trailing-edge pop-up for newly sent rows.
A focused conversational surface with visual tones, independent alignment, grouped messages, expandable content, and interactive link or button support.
A reader-aware conversation viewport that follows streamed output at the live edge and releases control when the reader moves away.
An auto-growing agent composer with prompt actions, model selection, keyboard submission, and animated send and stop states.
A stable response surface with completion actions, rendered content, and an expandable source summary.
A collapsible agent task plan with morphing status marks, a completion count, compact metadata, and smooth list updates.
A human-in-the-loop permission card for reviewing tool details, allowing once, remembering access, or denying execution.
A lightweight execution disclosure for syntax-highlighted terminal output and request responses that collapses into a compact completed state.