Anomaly has officially released OpenCode 2.0, three months after its public beta, introducing a shared background service that connects the open-source coding agent's terminal, web and desktop interfaces. The company announced the release in a blog post on Monday, describing it as a substantial overhaul informed by "millions of daily users" and designed to support a growing set of agent capabilities. The upgrade also includes a redesigned API, an embeddable SDK, and support for running subagents and shell commands in the background.

The most notable architectural change involves how OpenCode operates and how its various interfaces communicate. The background service now starts automatically as a single process on a developer's machine, allowing the terminal, browser and desktop interfaces to interact with the same underlying service. This setup lets developers launch a session in the terminal interface, continue it through the desktop or web app, and have their own scripts connect to the same service. Scripts using OpenCode's programmatic client can automatically find and connect to the local service, making it suitable for scheduled tasks, while developers can also expose the service to other machines and link it with external applications. Version 2.0 also changes how agents work with external tools through the Model Context Protocol, with a feature called Code Mode allowing agents to write JavaScript that calls multiple tools and processes results in a single execution. Built around a lightweight JavaScript interpreter, it turns MCP servers and OpenAPI endpoints into functions that agents can use programmatically, meaning an agent could pull bug reports from GitHub, extract details and create issues in Linear without separate model round trips for each step. Code Mode becomes the default for all MCP servers in the new version, with the company stating this approach makes extensive integrations practical. The release also brings background execution for both subagents and shell commands, letting developers interact with the primary agent while other operations run concurrently, such as delegating a code review to a subagent using a different AI model while working on the main task.

According to Anomaly, the previous API "grew haphazardly as we and the world collectively learned what agents are and should be," adding bluntly that "it was bad." The redesigned API gives agents greater control over their own environment, including the ability to create a Git worktree for a new task, transfer an existing session into that worktree, or identify and interrupt another session that has become stuck. The company also notes it has built more than 80 internal plugins covering most of version 2.0's functionality as part of an effort to improve the reliability and maintainability of its plugin API, while terminal interface plugins have moved from experimental to stable status. OpenCode's GitHub repository has accumulated over 212,000 stars and nearly 30,000 forks since its 2025 debut, gaining particular traction among developers seeking greater freedom over the AI models powering their coding tools.

The new embeddable SDK, distributed as @opencode/sdk, lets developers incorporate OpenCode directly into their own applications using the same underlying code as the CLI, enabling customized versions for servers or environments such as Cloudflare Durable Objects. However, there are compatibility implications for existing users: plugins developed for OpenCode 1.0 won't function with version 2.0, while integrations built around the previous server API will need updates. Existing users can upgrade to version 2.0 by running opencode upgrade --major from the command line or entering /update in the terminal interface, though according to the October 11 release notes, updates are no longer automatic and users must initiate the upgrade themselves. The shift to a client-server architecture with hot-reloading support positions OpenCode to handle more complex workflows without requiring developers to restart sessions when adding new capabilities. For teams building custom development environments or integrating AI agents into existing toolchains, the modular design may prove decisive in determining whether such tools can scale beyond prototype status.