Save your spot

Live eventOpen Bot: AI coworker

Copilot Kit Logo
  • Docs
  • Blog
  • AG-UI

    AG-UI (Agent–User Interaction) is the general-purpose, bi-directional connection between a user-facing application and any agentic backend.

  • Pricing
Get StartedTalk to an Engineer
AG-UI Streams

Conversations that pick up where your users left off.

With Intelligence’s AG-UI streams, users reconnect, catch up, and continue across devices. Messages, generative UI, tool activity, and app state are ready when they return.

Agent prompt

Paste into a coding agent running in your project folder.

Talk to an engineer
Any agent frameworkOpen protocolsSelf-hostableBuilt for your product

The whole conversation. Ready to continue.

Your framework manages the agent’s conversation and model context. Intelligence’s AG-UI streams deliver that interaction to your users, including messages, generative UI, tool activity, and application state.

Illustrative example: analyzing quarterly vendor spend. Framework capabilities vary.

Framework-only threads

The conversation stays in text.

User: Analyze our Q3 vendor spend
Agent:

Total Q3 vendor spend was $184,230 across 47 vendors. The largest were AWS at $52,400 (28.4%), Salesforce at $31,200 (16.9%), Datadog at $12,800 (6.9%), and Figma at $8,900 (4.8%).

Spend grew 12% quarter over quarter, driven primarily by infrastructure and observability. Would you like a further breakdown by category, month, or cost center?

User: Use the updated spreadsheet instead
Agent: I can’t open files here. Please paste the contents as text.
User: Anything we’re overpaying for?
Agent: I’d need the updated spend and usage data to identify unused licenses. Can you paste that here too?
Connection lostResponse interrupted. Start again to continue.

AG-UI stream

The conversation carries the experience.

User: Analyze our Q3 vendor spend
Agent:

Parsed 47 vendors. Here’s the breakdown:

Spend by vendor

Total vendor spend
$184,230
AWS
$52.4k
Salesforce
$31.2k
Datadog
$12.8k
Figma
$8.9k
Other
$78.9k
User: Use the updated spreadsheet insteadq3-vendors.xlsx2.1 MB
User: Anything we’re overpaying for?
Agent: Reparsed with the new sheet. Datadog looks high for your usage, and 3 licenses appear unused. Cutting those seats saves $4,100 per quarter. Drafting the savings memo.
Connection restoredThe response continues. Your file, chart, and context stay with the thread.

Generative UI

Keep charts, tables, forms, and interactive components with the conversation. Users return to the experience they were working in, with the context to continue.

Files and multimodality

Bring documents, images, and audio into the thread. Keep the material a user shares alongside the agent’s responses.

Tool activity

Preserve the actions behind an answer. Tool calls and results stay with the conversation so users can follow what happened.

Application state

Restore the state that gives a conversation meaning. The thread carries more than the visible messages when a user picks it back up.

Reconnect and catch up

An agent run can continue in the background after a browser disconnects while your runtime and agent stay running. Reconnect to catch up on the interaction.

Multiple devices

Open the same thread on web and mobile, even mid-stream. Let people continue their work wherever they need it.

Your conversations.
Richer experiences. Smarter agents.

Bring durable delivery and Automatic Learning to the conversations your agent already has, through the supported Intelligence integration. For historical conversations, use the existing Google ADK or LangGraph import guide for your deployment.

No thread provider yet? Use CopilotKit’s built-in thread store to save and resume conversations. Already have one? Keep it and layer on AG-UI streams and Automatic Learning.

01

Connect or start your threads

Connect your existing conversation history or start new threads with CopilotKit’s thread store. Choose which threads a Learning Space should learn from.

Supercharge your framework’s threads
02

Review what the agent learns

Automatic Learning finds patterns in collected threads and proposes reusable Skills. Review the evidence and approve what should be available to your agent.

03

Put approved Skills to work

Configure delivery once. Approved Skills become available in future interactions, and the agent decides when to use them.

Explore self-improving agents

Your agent stack.
Your product experience.

Intelligence’s AG-UI streams work through CopilotKit components and supported agent integrations. AG-UI defines the interaction events; Intelligence adds persistence, reconnection, and delivery.

Explore the Frontend SDKs

Self-hosted, with your database

Run Intelligence on your infrastructure and keep conversation data under your control. AG-UI Streams fit your product’s deployment and security requirements.

  • Bring your own database
  • Automatic migrations
  • Control over data residency and security policies
Explore self-hosted Intelligence

Managed in CopilotKit Cloud

Use AG-UI Streams without managing the underlying infrastructure. CopilotKit handles scaling, updates, and maintenance.

Explore plans

Add AG-UI Streams to your product.

Give your coding agent the setup prompt to connect your app, or bring existing conversations into Intelligence.

Agent prompt

Paste into a coding agent running in your project folder.

Talk to an engineer
Try Intelligence for free