CopilotKit
The frontend stack for agentic user experience — drop-in chat, generative UI, and shared state for React apps on any AG-UI backend
If you're a React team with an agent backend and you're hand-rolling streaming, state sync, and approval UI, CopilotKit removes months of plumbing. Named evidence: Rich Threads persistence, shared state, Human-in-the-loop primitives, generative UI across A2UI/Open-JSON-UI/MCP Apps, and documented paths to LangGraph, Claude Agent SDK, Google ADK, AWS Strands, Mastra, and PydanticAI. The August 2026 Channels SDK takes it into Slack and Microsoft Teams. Skip it if your frontend isn't React, or compare against Vercel AI SDK if you want lower-level primitives rather than a full agent-UX stack.
Verified 1d ago · liveness 87/100 · cite: rightaichoice.com/tools/copilotkit
- React teams shipping a copilot or assistant inside an existing product
- Developers who need agents to render real React components, not just text replies
- Products requiring human-in-the-loop approvals and edits on agent actions
- Teams on LangGraph, Claude Agent SDK, or Google ADK who want to keep their backend
- Non-React frontends — the runtime and primitives assume React
- Simple FAQ or scripted chatbots where a basic widget is enough
- No-code or low-code builders with no appetite for writing frontend code
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip CopilotKit if your frontend isn't React, or if your agent backend doesn't speak AG-UI and matches none of the partner integrations — you'd be writing an adapter before seeing any value.
The open-source React core, prebuilt chat components, headless UI, generative UI primitives, and AG-UI backend integrations are free, which makes it unusually cheap for a solo developer or small startup shipping a first copilot. Enterprise pricing is custom and gates the Enterprise Intelligence tier plus talk-to-an-engineer onboarding, so the cost jump happens when you need evaluation and rollout support for a production agent UX rather than at a usage threshold.
In short
CopilotKit — The frontend stack for agentic user experience — drop-in chat, generative UI, and shared state for React apps on any AG-UI backend. Best for React teams shipping a copilot or assistant inside an existing product, Developers who need agents to render real React components, not just text replies, Products requiring human-in-the-loop approvals and edits on agent actions. Free to use.
What's new in CopilotKit
Checked yesterdayAcross the latest 1 update: 1 changelog entry.
Viability Score
How well maintained and how widely used is CopilotKit? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: September 2026
How we score →Key Features
- CopilotChat, CopilotSidebar, and CopilotPopup drop-in chat surfaces
- Generative UI that renders real React components from agent output
- Headless UI for full control over every pixel
- Rich Threads persist messages, generative UI, and tool activity across sessions and devices
- Shared state between agent and application keeps conversations grounded in the current view
- Human-in-the-loop workflows for approvals, edits, and corrections
- Built-in in-process agent to start building without an external backend
- Any AG-UI compatible backend, including LangGraph, Google ADK, AWS Strands, and Mastra
- Claude Agent SDK integration with generative UI, shared state, HITL, sub-agents, and streaming
- WebMCP support on React 19 for exposing app functionality to agents
- Channels SDK for deploying agents into Slack and Microsoft Teams
- A2A and MCP handshake support for agentic interoperability
- Support for A2UI, Open-JSON-UI, and MCP Apps generative UI specifications
- Automatic Learning turns real usage into skills that improve the agent
- User Memories and sub-agents for richer agent capability
About CopilotKit
CopilotKit is a React frontend toolkit for building agent-powered product surfaces: chat, generative UI, shared state between your app and the agent, and human-in-the-loop approval flows. You bring an agent backend; CopilotKit provides CopilotChat, CopilotSidebar, and CopilotPopup for a working chat surface in minutes, headless UI if you want to control every pixel, and Rich Threads that persist messages, generative UI, and tool activity across sessions and devices. The state layer is the differentiator. Shared state keeps the agent grounded in what your user is actually looking at, while generative UI lets agents render real React components — not just text. Human-in-the-loop primitives let you require approvals and edits before agent actions land, and WebMCP on React 19 exposes app functionality to agents. Generative UI spec support spans A2UI, Open-JSON-UI, and MCP Apps, with A2A and MCP handshakes for interoperability. Backend choice stays open. The Built-in Agent gets you running immediately, and documented partner paths cover LangGraph (Python, FastAPI, TypeScript), LangChain Deep Agents, Google ADK, AWS Strands (Python and TypeScript), Mastra, Claude Agent SDK (Python and TypeScript), PydanticAI, Microsoft Agent Framework (.NET and Python), MS Agent Harness (.NET), AG2, Agno, LlamaIndex, and CrewAI Flows. Any backend that speaks AG-UI can plug in. CopilotKit has also pushed past the browser: the Channels SDK (announced August 2026) brings agents into Slack and Microsoft Teams, alongside the separate launches of Vendo and Open Bot. Automatic Learning turns real usage into skills that improve the agent, with User Memories and sub-agents adding further capability. It's for React teams shipping copilots, support assistants, and internal tools inside real products — not demo pages. You pay in React commitment, not framework lock-in.
Behind the Verdict
CopilotKit's bet is straightforward: most teams building agent features don't need another orchestration framework, they need the UI half of the problem solved. The library delivers that half with unusual breadth. Strengths start with the UI primitives. CopilotChat, CopilotSidebar, and CopilotPopup give you a complete surface out of the box, and headless UI lets you keep the runtime while owning every pixel. Generative UI is the part competitors underserve — letting agents render real React components rather than streamed markdown changes what you can build, and CopilotKit supports the major specs (A2UI, Open-JSON-UI, MCP Apps) instead of betting on one. Shared state is the quiet workhorse: it keeps multi-turn threads grounded in the current view, which is the difference between a chat bolted onto a dashboard and an assistant that knows what you're looking at. Human-in-the-loop primitives cover approvals and edits, which production teams need before letting an agent mutate data. Backend breadth is the second strength, and it's documented rather than claimed: LangGraph in Python, FastAPI, and TypeScript; Google ADK; AWS Strands; Mastra; Claude Agent SDK; PydanticAI; Microsoft Agent Framework; AG2; Agno; LlamaIndex; CrewAI Flows; LangChain Deep Agents. Any AG-UI backend works. Recent news extends the reach further — the Channels SDK (August 2026) puts agents in Slack and Microsoft Teams, and CopilotKit also launched Vendo and Open Bot. Weaknesses are real. The runtime and primitives assume React; non-React frontends are out of scope. The cloud tier (Enterprise Intelligence) is newer and thinner on community resources than the open-source core, and the docs assume you already know your agent backend — a real hurdle if you're choosing one. And if your agent stack doesn't speak AG-UI or match a partner integration, you're writing an adapter before you see value. Where it fits: React product teams with an existing agent backend, internal tools teams, and anyone whose agent needs to render interface rather than text. Where it doesn't: no-code builders, scripted FAQ chatbots, and teams with no orchestration layer and no plan to build one.
Researching CopilotKit? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Real-world workflow fit
Concrete scenarios for the personas CopilotKit actually fits — and what changes day-one when you adopt it.
Drop CopilotSidebar into an existing admin dashboard, point it at the Built-in Agent to prove the interaction, then swap in a LangGraph (TypeScript) backend over the AG-UI adapter.
Outcome: A working 'chat with your dashboard' surface in days rather than weeks, with the backend swappable without rewriting UI.
Keep the existing Claude Agent SDK (Python) backend and wire CopilotKit in for generative UI, shared state, and human-in-the-loop approvals on agent actions.
Outcome: Agents render real React components and route risky actions through approval, without replacing the orchestration layer.
Use the Channels SDK to bring an existing agent into Slack and Microsoft Teams for support and ops staff.
Outcome: The same agent reaches the channels where staff already work instead of requiring a separate web surface.
Use Cases
- Add a 'Chat with your dashboard' sidebar to a SaaS admin panel using CopilotSidebar.
- Let an agent render charts, tables, and widgets inside a canvas-style whiteboard app via generative UI.
- Build an assistant that executes UI actions (filter, export, create) on behalf of the user with human-in-the-loop approval.
- Ground a multi-turn support thread in the ticket the user is viewing using shared state.
- Deploy an internal agent into Slack or Microsoft Teams with the Channels SDK.
- Expose app functionality to agents through WebMCP on React 19.
- Sync agent conversations across a user's devices with Rich Threads.
- Prototype a multi-step copilot flow on the Built-in Agent before wiring a production backend.
Models Under the Hood
as of 2026-09-14
Limitations
- CopilotKit is explicitly a frontend stack for agentic user experience, so it targets React apps and requires a backend that speaks AG-UI (its own built-in in-process agent or a partner framework like LangGraph).
- Documentation is organized around specific agent backends, which assumes you have or are choosing one.
- It is an agent UX layer rather than a model abstraction, so provider-level model choices live in the backend, not here.
as of 2026-09-14
Verification history
We have re-verified CopilotKit 18 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
Showing the 6 most recent of 18 verification passes.
Free to cite with attribution — this page re-verifies continuously.
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Plans compared
For each published CopilotKit tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
Free
$0/mo
Ideal for
Solo React developer or small startup shipping a first copilot who is comfortable self-hosting on their own runtime.
What this tier adds
Free entry point: open-source React core, prebuilt chat components, headless and generative UI primitives, AG-UI backend integration, and the Free CopilotKit Intelligence tier.
Enterprise
Custom
Ideal for
Product teams moving an agent UX into production that need evaluation, rollout support, and deployment guidance.
What this tier adds
Adds talk-to-an-engineer onboarding, the Enterprise Intelligence tier, and evaluation and rollout support on top of the free open-source layer.
Where the pricing makes sense
The company stage and team size where CopilotKit's pricing actually pencils out — and where peers do it cheaper.
The open-source React core, prebuilt chat components, headless UI, generative UI primitives, and AG-UI backend integrations are free, which makes it unusually cheap for a solo developer or small startup shipping a first copilot. Enterprise pricing is custom and gates the Enterprise Intelligence tier plus talk-to-an-engineer onboarding, so the cost jump happens when you need evaluation and rollout support for a production agent UX rather than at a usage threshold.
Setup time & first value
How long it actually takes to get something useful out of CopilotKit — broken out by persona, not the marketing-page minute.
React developer with an existing AG-UI backend: first chat surface in under an hour using CopilotChat, with the Built-in Agent you can see an interaction in minutes. Non-React or non-AG-UI stack: budget days to write an adapter, plus extra time if you're still selecting an agent backend, since the docs assume you already know yours.
Switching to or from CopilotKit
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From Vercel AI SDK: replace hand-built chat and state plumbing with CopilotChat plus shared state, keeping your backend behind an AG-UI adapter.
- →From a hand-rolled React chat widget: swap the custom UI for CopilotChat or headless UI and inherit Rich Threads persistence and tool activity.
- →From a plain text chat box: move to generative UI so agents render React components instead of streaming markdown.
- →From a browser-only assistant: add the Channels SDK to reach Slack and Microsoft Teams without rebuilding the agent.
- ↗To Vercel AI SDK: rebuild the chat surface and state layer on lower-level primitives if you want fewer opinions and no AG-UI dependency.
- ↗To an AG-UI-native frontend you maintain: keep the same backend and replace the UI primitives.
- ↗To a non-React frontend stack: the CopilotKit runtime must be replaced, since the primitives assume React.
Integrations
Resources & Guides
- Quickstartdocs.copilotkit.ai
the frontend stack for agents
Connect any agent framework or model to your React app for chat, generative UI, canvas, and human-in-the-loop workflows.
- Tutorialdocs.copilotkit.ai
the frontend stack for agents
Connect any agent framework or model to your React app for chat, generative UI, canvas, and human-in-the-loop workflows.
- Conceptsdocs.copilotkit.ai
concepts
Docs, live demos, and integrations for CopilotKit.
- Learndocs.copilotkit.ai
concepts
Docs, live demos, and integrations for CopilotKit.
Tutorials & Learning
YouTube returned 6 videos for “CopilotKit”, and we withheld 6: 6 could not be judged, because “CopilotKit” is a single word that other videos use for other things. We are showing none, because we could not prove any of them are about CopilotKit.
Official links
Tools that pair well with CopilotKit
Common stack mates teams adopt alongside CopilotKit, with the specific reason each pairing earns its keep.
Featured Head-to-Head Comparisons
Alternatives to CopilotKit
View all
Frequently Asked Questions
Used CopilotKit? Help shape our editorial sentiment research.