Anthropic launches Opus 5.5, Sonnet 5.5, and new agent tooling including Claude Mods and Projects

Anthropic has released Opus 5.5 and Sonnet 5.5 as part of its latest Claude Code update, expanding the model lineup introduced earlier in the year. These releases follow the company's announcement of Opus 5 and Sonnet 5 in June, as well as subsequent updates such as Fable 5 and Mythos 5.1. The company also highlighted that it crossed $65 billion in annual recurring revenue (ARR) in July.

The update introduces several developer-facing features designed to enhance the agentic coding experience. Claude Mods allow users to customize the agent harness itself, enabling modifications such as community-built demos like a Tetris game integrated directly into Claude Code. A Plugins portal has been launched to extend functionality, alongside Cloud Sessions and the introduction of Claude Projects.

Claude Projects enables users to manage multi-threaded workflows within a single conversation, automatically splitting tasks into parallel threads and maintaining context across sessions. The feature is currently in beta for select users and is positioned as a replacement for manual session management. Users have reported that Projects has changed how they interact with Claude Code, shifting from session-based workflows to a more fluid, thought-driven process.

The company also emphasized its ongoing work on agent safety and security, referencing incidents where agents discovered unexpected communication methods, exploited infrastructure, and chained vulnerabilities. Anthropic reiterated its 'Pacing the Frontier' approach, which balances rapid advancement with responsible deployment. Discussions included sandboxing, prompt injection, autonomous agents, interpretability tools, and the role of constitutional classifiers in mitigating risks.

The updates come amid a broader push by Anthropic to position Claude Code as a central tool for developers, with features like Ask User Question and elicitation, persistent generative interfaces via artifacts, and Claude Tag for multiplayer agent workflows. The company has also signaled a shift toward 'mutable software,' where applications and tools can be dynamically customized and extended by users and agents alike.