claude-opus-4-5-migration

This migration tool streamlines the transition to Claude Opus 4.5 by automating the complex process of updating model identifiers, headers, and prompt structures across your development projects. Designed for software engineers and AI developers, it ensures your integration follows platform-specific best practices for Anthropic API, AWS Bedrock, Google Vertex AI, and Azure AI Foundry. The agent intelligently identifies existing model strings like claude-sonnet-4-20250514 or claude-opus-4-1-20250422 and replaces them with corresponding Opus 4.5 equivalents. Beyond simple string replacement, it manages environmental configurations, such as removing unsupported legacy beta headers and recommending the 'high' effort parameter for improved reasoning performance. It also provides targeted prompt adjustments to resolve common Opus 4.5 behavioral nuances, including tool overtriggering, over-engineering tendencies, and frontend design refinement, ensuring your AI agents remain stable and cost-effective after the upgrade.

• Automatically scans codebase for outdated model identifiers and API strings across multi-platform cloud deployments.

• Handles legacy beta header removal, specifically identifying and documenting incompatible 1M context beta headers.

• Offers optional, context-aware prompt snippets to mitigate tool overtriggering, excessive abstraction, and conservative code exploration.

• Refines language sensitivity by providing 'thinking' alternatives for non-extended thinking API requests.

• Provides a safe migration path by defaulting to string updates while keeping behavioral adjustments optional and user-driven.

• The agent is intended for use when upgrading existing LLM integrations; do not use it to migrate Haiku models which are explicitly excluded from this workflow.

• Ensure you have read access to the target files; the agent will modify existing source code files directly.

• For best results, use this agent in an environment where you can review diffs, such as within the Claude Code CLI or a git-enabled workspace.

• Always verify output against your specific platform's API requirements if utilizing custom infrastructure beyond standard providers.