16 components

AI Agents

Chat, tool-call and reasoning interfaces for AI agents, ready to own.

Prompt Input
An auto-resizing prompt box with attachments and animated submit states.3 variants
Agent is Performing Task
0/4
Read project files
Plan implementation
Write component code
Run tests
Agent Task List
A live checklist that shows an agent's plan ticking off as it works.2 variants
TSsum.ts
const sum = (a: number, b: number) => a + b console.info(sum(2, 3))
Code Block
Syntax-highlighted code with copy feedback and a tidy header.1 variant

TSformat-price.ts+1-1

export function formatPrice(cents: number) {
  return `$${(cents / 100).toFixed(2)}`
  return fmt.format(cents / 100)
}

Inline Diff

Inline additions and removals so users can review what an agent changed.1 variant

R
Refactor agent
Ready · 0.0s
Agent Run
A step-by-step run timeline with tool calls and approval gates.2 variants
Support agent · 5 steps, 4 connections
Next run: routine ticket
Ticket receivedInbox: support@TriggerIdle
Classify intentModel: router-v2AgentIdle
Page on-callPager: #sev-1AlertIdle
Draft replyModel: writer-v1AgentIdle
Send replyEmail: customerActionIdle
Press enter or space to select a node. You can then use the arrow keys to move the node around. Press delete to remove it and escape to cancel.
Press enter or space to select an edge. You can then press delete to remove it or escape to cancel.
Select a step to see its details
Agent Flow
A draggable node canvas that runs an agent workflow, built on React Flow.2 variants
Reading your question...
Dither Loader
Retro dithered loaders for thinking and generating states, plus a compact 3x3 cube.2 variants
According to recent studies, artificial intelligence has shown remarkable progress in natural language processing. The technology continues to evolve rapidly, with new breakthroughs being announced regularly .
Citations
Source chips and cards that show where an answer came from.1 variant
No models found.
Claude Opus 5.5Deepest reasoning for hard problems
Claude Sonnet 5.5Balanced speed and intelligence
Claude Haiku 4.5Instant answers, lowest cost
GPT-5Flagship general-purpose model
GPT-5 miniFast and affordable for everyday tasks
Gemini 2.5 ProLong context and strong multimodal
Gemini 2.5 FlashLow-latency, high-volume work
DeepSeek R1Open reasoning model
Llama 4 MaverickOpen weights, run anywhere
Reasoning effortClaude Opus 5.5
Model Selector
A model picker with a sliding highlight and capability hints.2 variants
Understanding the question
You want to know why consistent sleep helps memory — I'll look for studies and keep the answer practical.
Reasoning Trail
A collapsible trail of the reasoning steps behind a response.1 variant
A few quick questions
Step 1 of 4
What are you trying to build?
This helps me pick the right starting point.
AI Questionnaire
Step-through questions an agent asks before it starts a task.1 variant
Dither Voice Orb
A tappable dithered orb that swells and ripples with the voice level.1 variant
ComposeRenderRefineUpscale
0%
“a misty mountain range under a full moon”
1024 × 1024
Image Generation
Prompt-to-image cards with progressive reveal while generating.1 variant
AI Chart Card
A chart an agent drew from a prompt, with a generating state and a written takeaway.1 variant
Thinking Text
Shimmering status text for when an agent is thinking.1 variant
Searching
Notion – Q3 funnel report
Agent Thinking
A collapsible Thinking header with live timer, source chips and step-by-step reasoning.1 variant

AI Agents

What are you trying to build?