About agentscreens

agentscreens is a curated reference for UI patterns in production AI agents — how tool calls, streaming, citations, plans, errors, code execution, and other agent-specific UX moments are actually rendered by the products people use every day.

Pattern libraries exist for prompts. They don't really exist for what builders actually struggle with: how do I show a tool call? How do I render streaming with citations? What does a good plan view look like? How should errors fail gracefully? This is a place to look those up.

It's open source under the MIT license. Submissions are welcome via PR or the submit page. The site is deliberately simple — static, fast, no login required to browse.

Built by 0xdeadd.

What gets curated

• Tool calls — invocation summaries, argument viewers, result rendering
• Streaming — typing animations, partial responses, cursor styles
• Citations — inline references, source panels, hover previews
• Plans — TODO lists, hierarchical task trees, status indicators
• Errors — failure banners, retry affordances, refusal styling
• Code execution — terminal output, build logs, live previews
• Approvals — human-in-the-loop confirmations for sensitive actions
• Loading— "thinking", "searching" states, shimmer placeholders
• Memory— surfacing remembered context, "saved to memory" pills
• Multi-agent — sub-agent dispatch, parallel work, handoffs