Vercel AI SDK Adds Deep Agents and OpenCode: Unifies Development with Code Agents

What Happened?

Vercel has announced the addition of two new adapters to its AI SDK Harness: Deep Agents and OpenCode. These adapters allow running code agent runtimes through a unified interface, so developers can switch runtimes without altering their application code. Both run inside a Vercel sandbox, providing a secure and isolated environment.

The @ai-sdk/harness-deepagents adapter adapts the LangChain runtime 'deepagents', and includes built-in file and shell tools, skills, host tools, multi-turn sessions, attach and resume capability, and tool approval. Meanwhile, @ai-sdk/harness-opencode starts a real OpenCode server inside the sandbox via @opencode-ai/sdk and streams its session events through the harness. It exposes OpenCode's built-in tools, supports both built-in and host tool approval, and allows selecting the model, provider, and reasoning variant.

With these, the full list of supported harnesses is: Claude Code, Codex, Deep Agents, OpenCode, and Pi, with more expected in the future. According to Vercel's official blog, the goal is to offer a consistent experience for running code agents, regardless of the underlying runtime.

Why Is This Important?

Unifying multiple code agent runtimes under a single interface reduces development friction. Teams can test different agents without rewriting integration logic, accelerating experimentation and adoption of new capabilities. Additionally, running in a sandbox minimizes security risks by allowing arbitrary code execution in a controlled environment. This is especially relevant for applications requiring autonomous agents that interact with the file system or execute commands.

From a market perspective, Vercel strengthens its position as a preferred platform for AI development, competing with solutions like LangChain and offering an abstraction layer that simplifies the complexity of code agents. For developers, it means less time on setup and more on business logic. This move aligns with the industry trend toward standardizing interfaces for AI agents, similar to how REST APIs standardized communication between services.

Consequences and Projections

This move could accelerate the adoption of code agents in production by lowering the technical and security barrier. We are likely to see an increase in applications using agents for tasks such as code generation, automated debugging, or pull request review. It could also incentivize other runtime providers to offer adapters for Vercel's AI SDK, expanding the ecosystem. For example, if popular runtimes like AutoGPT or AgentGPT create their own harnesses, interoperability would increase significantly.

However, sandbox execution may introduce latency and resource limitations. Developers will need to evaluate whether performance meets their application requirements. Additionally, dependency on a specific vendor (Vercel) could be a consideration for teams seeking portability. In the long term, Vercel could become the central orchestrator of AI agents, similar to what it did with serverless deployment.

What Should Readers Know?

If you already use Vercel's AI SDK, you can now integrate Deep Agents and OpenCode by simply installing the corresponding npm packages and configuring the harness. The official documentation provides quick start guides. For those not using Vercel, this news indicates where the industry is heading: toward standardizing interfaces for AI agents.

It's important to note that while the adapters are new, the underlying runtimes (Deep Agents and OpenCode) already have their own communities and use cases. This integration does not replace those tools but makes them more accessible from the Vercel ecosystem. Deep Agents, for example, is a LangChain runtime that allows creating agents with complex skills, while OpenCode is an open-source agent server that supports multiple models and providers.

In summary, Vercel takes another step toward becoming the orchestration hub for AI agents, and developers should explore these capabilities to stay competitive. The combination of a secure sandbox, a unified interface, and the flexibility to choose among multiple runtimes positions Vercel as a key player in the next wave of AI-assisted development.