context-mode

Use context-mode to route large log files through indexed, sandboxed execution that summarizes, filters, and extracts targeted findings like error counts and patterns without loading raw content into the agent context.

The best way to extract findings from large build outputs is using context-mode analysis tools to parse and summarize the results. This approach yields explicit, actionable findings like error counts and affected files instead of returning the complete, overwhelming command output.

Yes, you can process Playwright browser snapshots by saving them to files first. Context-mode handles these saved browser snapshots through indexed file workflows, allowing you to search and extract targeted findings without injecting the raw DOM into the agent context.

Yes, context-mode works with MCP tool results and API responses. It routes sizable MCP responses through context-mode analysis tools to parse, filter, and summarize the data, preventing large API outputs from overwhelming the agent context.

You can index and search external documentation by applying context-mode workflows that use sandboxed execution and file-based search. This allows you to query and extract specific information from indexed files without loading the entire documentation into context.