Microsoft Foundry adds Claude Opus 5.5 for long-running coding and knowledge work

Anthropic's Claude Opus 5.5, designed for everyday complex work, is now accessible to developers and enterprises building AI applications and agents on Microsoft Foundry. It aims to advance agentic coding, knowledge work, and long-running tasks.

The model introduces improvements to agentic communication, enabling it to surface key information like actions taken, findings, decisions, input needs, and task outcomes. This helps users understand the model's process and results without inspecting every step.

Anthropic is introducing beta API capabilities alongside Opus 5.5 to support long-running agents, including asynchronous and keep-tail compaction, and the ability to change tools during a conversation while preserving context and prompt caching.

Claude Opus 5.5 is optimized for multi-step tasks like building features across a codebase, debugging, refactoring, and extensive knowledge work such as synthesizing information from multiple sources. It aims to deliver more usable outputs requiring less editing before sharing.

Claude Opus 5.5 automatically determines the reasoning effort required for a task, allowing it to adapt its analysis complexity. It also offers lower per-token prices and much cheaper cache reads, reducing costs for developers.

Claude Opus 5.5 incorporates expanded safeguards, utilizing safety classifiers similar to Claude Fable 5.1 for areas like cybersecurity and AI development. Certain higher-risk or dual-use requests may be handled by other Claude models with appropriate safeguards.