CopilotKit releases v1.59.2 with threads, Vue packages, and React Native SDK

• CopilotKit's launch post says v1.59.2 adds threads, Vue packages, a React Native SDK, and v2 rendering hooks on top of the project's AG-UI-based full-stack agent UI stack.
• According to the Threads docs, the new thread layer is built around a useThreads hook with realtime sync over WebSocket, plus server-side history so conversations can resume across devices.
• CopilotKit's React Native announcement and the React Native docs frame mobile support as headless first: you get the agent hooks and runtime integration, but you build the UI yourself.
• The AG-UI docs describe the protocol under CopilotKit as a lightweight event-based standard for wiring agents to user-facing apps, which is why the launch post can list LangGraph, Mastra, ADK, PydanticAI, LlamaIndex, CrewAI, AG2, and Claude Agents in the same compatibility breath.
• CopilotKit's chat-surface teaser pushes the roadmap past the browser, with early-access generative UI demos running inside Slack, Teams, Google Chat, and Discord.

You can browse the repo, skim the AG-UI protocol docs, and read the new threads and React Native docs without hunting through a release thread. One useful wrinkle: the Vue work looks bigger than a wrapper package, because PR #4400 describes a production-ready Vue 3 port with parity targets for chat components, agent hooks, and A2UI rendering.

CopilotKit is positioning itself less as a chat widget and more as an agent frontend harness. The launch post groups shared state, prebuilt and headless chat UI, frontend tool calls, generative UI, human-in-the-loop flows, and MCP apps into one stack, while the repo README describes the project as a frontend stack for agents and generative UI.

The AG-UI docs say the common layer is an event-based protocol for connecting agents to user-facing applications. That protocol abstraction is what lets CopilotKit claim compatibility across LangGraph, Mastra, ADK, Strands, MAF, PydanticAI, Agno, LlamaIndex, CrewAI, AG2, and Claude Agents in the same announcement.

Vue is part of that expansion too. The release summary mentions new Vue packages, and PR #4400 says @copilotkit/vue aims for Vue 3 parity with provider context, chat and sidebar components, useAgent, useThreads, useFrontendTool, and A2UI rendering.

Threads are the most concrete product addition in v1.59.2. The Threads docs say useThreads can list, create, rename, archive, and delete conversations, with metadata updates pushed to connected clients in realtime over WebSocket.

That same doc adds the persistence model the tweet does not spell out: conversation history is stored server-side in CopilotKit's Enterprise Intelligence Platform, so users can close the browser and resume later. That makes the new thread support closer to an app-level conversation system than a local chat history sidebar.

The mobile story is more specific than the launch thread's one-line bullet. The React Native docs say @copilotkit/react-native is a lightweight, headless wrapper around CopilotKit's React hooks, built for React Native 0.70+ and Node.js 20+, with a required runtime endpoint and Hermes polyfills loaded before other code.

The main tradeoff is explicit in the docs: there are no prebuilt UI components, so developers render their own chat surface with React Native primitives. The hooks are the familiar CopilotKit pieces, including useAgent, useFrontendTool, useHumanInTheLoop, and useCopilotKit, which lines up with CopilotKit's follow-up post describing mobile support as the same building blocks, now on iOS and Android.

CopilotKit's more interesting near-term reveal is that its generative UI work is already being adapted to Slack, Teams, Google Chat, Discord, and similar chat surfaces. The teaser post opens early access before a walkthrough on the implementation challenges.

That matters because CopilotKit's own Generative UI overview breaks the product into several rendering primitives, not just chat:

• Components as tools
• Tool call rendering
• Shared state
• Frontend actions
• Layout generation from a catalog
• MCP apps

CopilotKit's A2UI PDF demo shows what that looks like in practice on the web: upload a PDF, ask a question, and the agent selects from 21 React building blocks to assemble the interface on the fly. The chat-surface post suggests the same model is now being pushed into messaging clients, which is a much stranger target than yet another in-app copilot panel.