Configure options below — the preview updates in real time.

Cascade (AI Assistant)

Primary settings for controlling the Cascade AI agent behavior.

Allow Cascade in BackgroundAllows the Cascade AI agent to continue processing requests, generating code, and executing terminal commands even when the editor loses focus.
Auto Generate MemoriesAllows Windsurf to automatically extract and save project context, coding preferences, and rules from your conversations into persistent memory banks.
Auto Open Edited FilesAutomatically opens files in the editor tabs whenever Cascade generates or modifies them in the background.
Cascade Auto Fix LintsAllows Cascade to automatically parse linting errors and stylistic issues and suggest or apply fixes.
Cascade Gitignore AccessGrants the Cascade AI access to read and edit files that are ordinarily hidden/ignored by your project's .gitignore rules.
Enable Cascade Web ToolsEnables Cascade to independently search the web, fetch documentation, and browse URLs. Note: Consumes more AI credits.

Controls whether Cascade can automatically execute commands in the integrated terminal. Four levels from most restrictive to least restrictive.

Commands matching entries in this list will be auto-executed without user confirmation.

Commands matching entries in this list will always require user approval before execution.

Enable Sounds for CascadePlays audio cues/sound effects when the AI finishes a long-running generation or background task.

Defines the maximum number of files Windsurf will index in your workspace for the AI to reason over.

Remember Last Model SelectionPersists the last selected AI model (e.g., Claude 3.5 Sonnet, GPT-4o) across editor reloads.
Open Recent ConversationAutomatically loads the last active Cascade chat thread when the IDE is opened.
Explain and Fix in Current ConversationWhen using Explain & Fix on a diagnostic, continue in the current Cascade conversation instead of starting a new one.

Overrides the font size specifically for text inside the Cascade chat panel.

Autocomplete & Supercomplete

Settings for inline AI code completion and Supercomplete predictions.

Enable AutocompleteMaster switch to enable or disable Codeium's inline AI code completion suggestions.
Enable SupercompleteToggles the Supercomplete engine, which predicts broader multi-line intents and jumps ahead across multiple files.
Enable Tab to JumpEnables the Tab to Jump feature, allowing you to press Tab to jump to the next predicted edit location across files.

Sets the processing speed target for inline completions. The "fast" profile routes to lower-latency models.

Extension Marketplace Overrides

URLs for the extension marketplace backend and detail pages.

The backend URL used to fetch IDE extensions. Can be overridden to the standard Microsoft VS Code marketplace.

The base URL pattern used to display extension details pages.

Project-Level AI Rules

Workspace-level instructions that Windsurf's AI reads on every request. Rules are defined as Markdown files in .windsurf/rules/ directory.

Markdown instructions placed in .windsurf/rules/ directory at the root of your project. Windsurf's AI reads these on every request to adhere to project-specific formatting, architecture rules, and preferred libraries. Supports YAML frontmatter with trigger modes: always_on, model_decision, glob, manual.

Uses .gitignore syntax. Windsurf will completely ignore any files matching these patterns when calculating context or indexing the codebase.