CopilotKit v1.50 is now available. This page gives an overview of what is new, what is coming, and what just stays the same.

Getting Started

Set your CopilotKit version to:

npm install @copilotkit/react-core@latest @copilotkit/react-ui@latest @copilotkit/runtime@latest

---

Overview

CopilotKit v1.50 is a major update. It includes many highly sought after new features, under the hood improvements and simplifications, and extensive improvements to existing core elements:

Feature	Description
useAgent	Superset of useCoAgent with more control, including: Shared State, Time Travel, Multi-Agent Execution and Agent Mutual Awareness
Threads and Persistence	Multiple, long running and resumable, agent conversations
New Design System	Deeper UX customization
Direct to LLM Improvements	Shared state & Additional supported LLMs
Zod schema support	Type safe compilation and runtime
Simplified Infrastructure	Removal of GraphQL
Backwards Compatibility	All CopilotKit v 1.10 compatible apps can be updated to v 1.50 with no required coding changes

---

Backwards Compatibility

Everything you’re already using continues to work. v1.50 is fully backwards compatible with client code using v1.10. In fact, as a first step in this pre-release period, we would love for you to start with just rebuilding, and let us know if you run into any problems.

V2 interfaces#

As we said above, all of the v1 (or unversioned) hooks continue to work with v1.50. To take advantage of new hooks you’ll use those marked as /v2 (shown below).

You can opt into any of the new developer interfaces or keep using the old ones. Mixing is allowed in the following configurations:

• New Chat component + old hooks

• Old components + new hooks

• New v2 APIs are exposed under a subpath:

---

useAgent

The v2 hook useAgent is a proper superset of useCoAgent which gives more control over the agent connection, including:

Shared state#

const { agent } = useAgent({ agentId: "my-agent" });

agent.state
agent.setState

Time travel#

const { agent } = useAgent({ agentId: "my-agent" });

agent.setMessages()

Multi-agent execution#

const { agent: langgraph } = useAgent({ agentId: "langgraph" });
const { agent: pydantic } = useAgent({ agentId: "pydantic" });

[langgraph, pydantic].forEach((agent) => {
  agent.addMessage({ id: crypto.randomUUID(), role: "user", content: message });
  agent.runAgent();
});

Agent Mutual Awareness#

You can even make agents aware of each other!

const { agent: langgraph } = useAgent({ agentId: "langgraph" });
const { agent: pydantic } = useAgent({ agentId: "pydantic" });

langgraph.setMessages(pydantic.messages)
pydantic.setMessages(langgraph.messages)

---

Coming Soon#

v1.50 is available now, with useAgent fully ready. The other features (listed below) will be added as they come out in the near future.

Threads & Persistence#

v1.50 delivers support for threaded conversations, with built in thread storage and connection components. Support for development with InMemory or SQLite is provided out of the box.

Thread storage

• For development: InMemory or SQLite out of the box

  
  const runtime = new CopilotRuntime({
    agents: {
      default: agent,
    },
    runner: new InMemoryAgentRunner(), // new abstraction for persistence
  });
  

• For production

  • Coming soon! - Use Copilot Cloud (managed) or Copilot Enterprise (self hosted) to persist threads to your own databases, get automatic stream reconnection, in-built product analytics.
    • 👉 Join the Enterprise Early Access waitlist!

Thread connection

• Loading a thread

  
  <CopilotChat threadId="xyz" /> // connects to runtime's runner
  

• Stream reconnection

  
  <CopilotChat threadId="xyz" /> // will reconnect if page is reloaded
  

---

New Design System#

You can continue to use the design system in the v1 interfaces, but v2 provides A full UI overhaul with deeper customization:

• New CopilotChat, CopilotSidebar, CopilotPopup
  • Just an import away at @copilotkit/react-core/v2
  • New slot system for flexible overrides
  • Much better out-of-the-box markdown rendering
• Suggestions API that works cleanly with agents

---

Zod Schema Support#

Many CopilotKit hooks now take Zod schemas, enforcing type safety at both compile and run time.

• useHumanInTheLoop - now takes Zod schemas
• useFrontendTool - now takes Zod schemas
• useRenderToolCall - new dev-experience principles

---

Simplified Infrastructure#

Earlier versions of CopilotKit required GraphQL installation. That is completely removed by the new architecture.

---

Direct to LLM Improvements#

Our direct to LLM integration now supports more models and implements shared state.

BasicAgent - Now with broader model support

const runtime = new CopilotRuntime({
  agents: {
    myBasicAgent: new BasicAgent({
		  model: "openai/gpt-5.4", // much broader model support
		  prompt: "You are a helpful AI assistant.",
		  temperature: 0.7,
		});
  },
});

Shared state support

const { agent } = useAgent({ agentId: "myBasicAgent" });
const { state, setState } = agent;

---

Sharing Feedback and Getting Help

Find an issue? We'd Love Your Feedback#

If you run into anything unexpected:

→ Please raise an issue in the CopilotKit Discord support channel

Your feedback is incredibly valuable at this stage.

Does Your Team Need Help?#

If you’re using CopilotKit in a production product or business, we’d love to help.

You can schedule time with the team here:

👉 Scheduling link