
CopilotKit
Put AI agents inside your product, in React, Slack and Teams
About CopilotKit
CopilotKit is for the moment your agent works and you have to put it in front of users. That turns out to be a substantial engineering problem of its own: streaming state into a UI, letting the interface render what the agent is doing, allowing a user to interrupt or correct it, and doing all of that again for every surface where people expect to find it.
The project's answer is AG-UI, a general-purpose bi-directional protocol between a user-facing application and any agentic backend. The value of a protocol rather than a widget library is that the agent and the interface stop being one codebase, so the same agent can appear in your React app, in Slack and in Microsoft Teams without three implementations diverging. Generative UI goes further, letting the agent render actual interface components rather than only text, which is the difference between a chatbot describing a table and an application showing one. It is self-hostable and positioned at enterprise scale, having raised $27M. Two honest caveats. The pricing ladder climbs steeply, from free to $39 a month to $100 per seat, so model your seat count before assuming the middle tier is where you land. And adopting a protocol is a deeper commitment than adopting a component library, which cuts both ways: harder to leave, but it is the reason the multi-surface promise works at all.
You connect your agentic backend, whatever framework it runs on, through AG-UI. The frontend then receives a structured, bi-directional stream rather than a text firehose, so the interface can display progress, intermediate state and results as they happen, and send user input back mid-run. Because the connection is a protocol rather than a bundled runtime, the same backend serves a React application, a Slack workspace and Microsoft Teams. Generative UI lets the agent choose and populate real components instead of returning prose that your frontend has to parse. Self-hosting is supported for organisations that cannot route this through a vendor.
- •AG-UI Protocol - A bi-directional standard between any agentic backend and any user-facing app
- •One Agent, Three Surfaces - React, Slack and Microsoft Teams from a single implementation
- •Generative UI - The agent renders real interface components rather than only text
- •Framework Agnostic Backend - Works with the agent stack you already chose, including the Claude Agent SDK
- •Self-Hostable - Deployable inside your own infrastructure rather than only as a managed service
- •Free Tier to Build On - Development is free; cost arrives when you run at scale
Product teams whose agent has to live inside an application rather than in a separate chat window, and especially anyone who needs it in Slack and Teams as well as their own UI. The protocol approach pays off precisely when there is more than one surface, and is over-engineering when there is exactly one. Enterprises get self-hosting and governance controls. If you are building an internal tool for a handful of people, a plain chat interface will get you there faster and cheaper.
Pricing
$0 - $100/mo
- Developer$0/mo
- Pro$39/mo
- Team$100/mo
- EnterpriseContact sales
From the vendor pricing page, 2026-09-21













