Windsurf is an AI IDE published by Codeium. It is one of the five tools Rule Taxonomy and Evolution in AI IDEs: A Mining and Survey Study selected for study on the basis that each lets developers explicitly define AI IDE Rules the IDE must follow during code generation and chat interactions. That study records its release date as 13 November 2024, per the tool’s official changelog.
Capabilities
Windsurf’s rule files live under .windsurf/rules/ in the project. It also supported a single-file .windsurfrules format, which was later superseded by the directory-based mechanism. Beyond the location and the fact that
rules are honoured during generation and chat, the source used here characterises the mechanism
generically across all five tools rather than describing Windsurf’s own implementation of it — see
AI IDE Rules for the shared account.
Adoption & Ecosystem
Two figures from that study bear on adoption, and they measure different things. In its repository mining, an initial search returned 3684 candidate Windsurf projects; after keyword and rule-file filtering and manual inspection, 5 remained in the final dataset of 83 projects declared as built with an AI IDE. In its practitioner survey, 23 of 99 respondents reported currently using Windsurf for development. The mining figure reflects how many public projects both use the tool and say so in their README or description, which the study notes undercounts projects that use an AI IDE without declaring it; the survey figure reflects self-reported use among developers who had committed changes to rule files.
A practitioner’s account of using it, Assessing internal quality while coding with an agent, describes adding a feature to an existing Swift Mac application with Windsurf and Sonnet 3.5, asking it for a plan for each chunk of work before the implementation. Its author found that the combination sped up writing code but required careful planning with prompts and constant switching between Windsurf and Xcode for building, testing and debugging, that the generated code had significant quality issues, and that the agent tended to get stuck trying to fix a problem — so that on the whole he did not feel he was getting much out of it. In the same account, a later attempt with Claude Code and Sonnet 4.5 went better enough for him to adopt that tool regularly.