Cursor is now available in the AI SDK harness layer
Vercel’s AI SDK harness layer now includes Cursor via the @ai-sdk/harness-cursor adapter, enabling developers to switch between multiple coding agents without modifying application code.
Vercel has introduced support for Cursor within its AI SDK harness layer through the new @ai-sdk/harness-cursor adapter. This adapter integrates Cursor into the existing HarnessAgent interface, allowing applications to run multiple coding agents—including Cursor, Claude Code, and Cline—through a unified protocol. The change eliminates the need to rewrite application code when switching between agents, streamlining development workflows. Developers can now leverage Cursor’s capabilities alongside other supported agents without additional configuration overhead.
The integration relies on the Agent Client Protocol (ACP) via the @ai-sdk/harness-acp package, which serves as the intermediary between Cursor and the HarnessAgent interface. This protocol standardizes communication across agents, ensuring compatibility and reducing fragmentation in tooling. Vercel’s announcement follows the expansion of supported harnesses, which now includes Codex, Deep Agents, Grok Build, OpenCode, and Pi. The company has indicated that additional agents will be added in future updates.
To facilitate adoption, Vercel has published documentation for the Cursor harness, outlining setup instructions and best practices for integration. The documentation covers configuration steps, required dependencies, and examples of how to initialize and manage the Cursor agent within the harness layer. Developers are advised to review the guide to ensure proper implementation and compatibility with their existing workflows.
The addition of Cursor to the AI SDK harness layer reflects Vercel’s ongoing effort to provide a flexible and extensible environment for AI-driven development tools. By supporting multiple agents through a single interface, the company aims to simplify experimentation and deployment for teams working with diverse coding assistants. The move aligns with broader industry trends toward modular and interoperable AI tooling in software development.