Documentation Index
Fetch the complete documentation index at: https://docs.devin.ai/llms.txt
Use this file to discover all available pages before exploring further.
Windsurf is now Devin Desktop.
This release updates the bundled Devin Local agent to 2026.5.26. Highlights are below — see the changelog for the full list of changes.
New features
- Devin Local is now aware of the files you have open in the editor as part of its context.
- When prompted for an MCP tool permission in Devin Local, two additional server-level options are now offered: approve all tools on the server for the current session, or permanently.
Bug fixes
- “Always Allow” permission grants in Devin Local now persist across sessions.
- Image attachments in Devin Local now show the correct warning when the selected model does not support images.
Bug Fixes and Improvements
- Fixed availability issues with swe-check model for some users.
- Improved terminal processing performance
- Restored conversation sharing
Devin Review & Quick Review
All Windsurf IDE users now have access to Devin Review and Quick Review with your existing subscription.Agent Command Center
- Added list display option for the agent inbox
- Improved sessions sidebar sorting and filtering
- Performance improvements for loading and switching sessions
Windows updates
We have fixed a bug preventing updates for some users on Windows. To upgrade to this version, you may need to:- Wait for the update to download.
- Once it is downloaded, open Windows Task Manager and close all
devin.exeprocesses - Proceed with installing the update and restarting Windsurf
Bug fixes and improvements
- Fixed bugs with some MCP servers
- Improved reliability of Devin Local agent
Settings Improvements
- Settings now open in a dedicated tab instead of a modal, making it easier to navigate and find what you need
- Added a searchable sidebar to quickly locate settings across all categories
- Consolidated the settings gear into the global activity dropdown for a cleaner titlebar
Agent Command Center Improvements
- Added button to rename agent sessions directly from the sidebar
- Added space @-mentions — you can now reference entire spaces in Cascade conversations
- Added automatic context sharing across sessions in a space
- Model picker search now surfaces the best-matching variant per model family
- Improved at-mention loading with incremental results instead of waiting for all categories
- Tab completions now respect your
files.excludesettings - The toolbar footer now works like the Cascade footer with accept/reject buttons tied to diff zones
- File paths and selection ranges are now included in @-mentions
- Session PRs are now rendered in the Devin Cloud sidebar item
- Fixed UI freeze that could occur when pasting large text into Cascade
Bug Fixes
- Fixed scroll position not being preserved when switching between editor tabs
- Fixed sign-in / sign-out state showing contradictory status after a failed authentication
- Fixed titlebar items disappearing in narrow windows
- Fixed stale session data showing in the sidebar
Devin for Terminal
- Updated Devin Local agent to respect existing Windsurf proxy settings
- Added context window indicator for Devin Local agent
- Improved revert support for Devin Local agent
- Enabled mentioning terminal content in Devin Local sessions
- Added new keyboard shortcuts for accepting permission requests from Devin Local agent
Bug Fixes and Improvements
- Added support for server-driven extension deny lists
- Simplified file drag & drop support in Agent Command Center
Devin for Terminal
Devin is now available for Terminal. All Windsurf users can use this new CLI agent with your existing subscription.- Runs on your machine — Optimized for interactive work, with full access to your codebase, tools, and environment.
- Hand off to the cloud — Seamless hand off to Devin in the cloud, with its own VM, testing, video recordings, autofix and more. Come back to a finished PR.
- Multi-model — All of your favorite frontier models in one place, including Opus 4.7, GPT-5.5, and SWE-1.6.
- Fast — Written in Rust and so performant that the binary can run on an original VT100.
Devin Agent in Windsurf
You can also enable the new Devin Local agent in Windsurf. This is the same agent harness used on the terminal and sessions can be accessed from both Windsurf and the CLI.In our testing, it’s up to 30% more token-efficient than the existing Cascade agent.Additional Changes
- Improved search subtitle layout during streaming
- Fixed file drag-and-drop in agent window Cascade tabs
- Fixed edit rendering issues in Devin Local on Windows
- Fixed Go to Line/Column keybinding on Windows/Linux (Ctrl+Shift+G)
- Stability and performance improvements
Agent Command Center
- Added workspace and session filters to the spaces sidebar
- Improved unread indicators for sessions with mark read/unread functionality
- You can now drag sessions onto kanban cards to create Spaces
- Added “Put Devin to sleep” tab context menu action
- Improved Cmd+F find-in-chat for agent sessions
- Hand off plans from Cascade to Devin Cloud with one click
Bug Fixes
- Fixed Cmd+P flicker when focused on cascade in agent mode
- Fixed Cmd+N to open untitled file when editor surface is focused in agent window
- Fixed Ctrl+Shift+M microphone keybinding in agent window mode
- Fixed agent sidebar highlight when focus moves to related panes
- Skip button now skips one question at a time in Devin Cloud sessions
- MCP registry now paginates instead of requesting limit=1000
- Honor HTTP_PROXY for Windsurf Remote-SSH language server
- Stop Devin for Terminal from spawning console windows on Windows
- Prevent Cascade input re-renders causing typing lag
Bug Fixes and Improvements
- Fixed OAuth authentication issues for some MCP servers
Bug Fixes and Improvements
- Fixed a regression with OAuth integration for some MCP servers
- Improved reliability of Devin Cloud connections in Windsurf
Bug Fixes and Improvements
- Various bug fixes and performance improvements for improving Windsurf 2.0 auth experience.
- Fixed bug with spawning Terminal sessions on Windows
Windsurf 2.0
Read the full announcement here.Devin in Windsurf
- Devin cloud agent available directly inside Windsurf, included with every self-serve plan
- Delegate tasks from a local session to Devin with one click; Devin runs on its own VM
- Review Devin changes and test results without leaving the editor
- Billing is based on your existing quota and extra usage, with up to 50 USD in extra usage added for launching your first Devin Cloud session.

Agent Command Center
- New Kanban-style view showing all local and cloud agent sessions, organized by status
- Group agent sessions, PRs, files, and context into task-level Spaces
- Switch between Spaces to switch between tasks
Additional Changes
- Refined Windsurf Browser with toolbar integration and Cascade tool for reading page contents
- Sped up initial load times for the Cascade sidebar
- Improved .gitignore and .codeiumignore handling across the product
- Stability improvements for remote extensions (WSL, SSH, Dev Containers)
- Performance improvements for typing in large active diff zones
Bug Fixes and Improvements
- Fixed resource parameter support for OAuth connections to MCP servers
- Enhanced MCP registry support
Adaptive Fix
We fixed a bug with the adaptive model router which prevented switching models after the first request.All users who encountered the bug have had quota reset and overage restored.Introducing Adaptive
We’ve made several model packaging changes, with more info here.Adaptive Model Router
A new Adaptive model option is now available in the model picker. Adaptive intelligently selects the best model for each task, helping you make your quota last longer by avoiding overuse of premium models.- Availability: Now available to all self-serve users on Pro, Max, and Teams plans.
- Dynamic model selection - Automatically chooses the right underlying model for your task while drawing down quota at a fixed per-token rate.
- Extra usage promo - Beyond your quota, extra usage is offered at 0.50 USD per 1M input tokens, 2.00 USD per 1M output tokens, and 0.10 USD per 1M cache read tokens for the next 2 weeks.
Token Tracking in Response Card
- Added token usage tracking to the response card. You can now see a breakdown of input tokens, output tokens, and cached input tokens for each response directly in the chat panel.
- Enhanced the context window indicator to show when the prompt cache expires, giving you better visibility into cache utilization.
Bug Fixes and Improvements
- Fixed keyboard input issues with the built-in terminal on Windows
- Improved cost sorting for token-based plans in the model picker
- Clarified which models are available only on pro plans
- Repaired devcontainer support on RHEL 8
A release was made.
A release was made.
A release was made.
Bug Fixes and Improvements
- Various bug fixes and performance improvements
Cascade
- Fix dangling Diff Zones related to Jupyter Notebooks
- Improve Jupyter Notebook performance when running on WSL
Cascade
- Improve Cascade UI rendering performance
- Fix Cascade agent crashes under certain conditions
- Improve notifications for model price changes
- Add support for loading SKILL.md files from the
.windsurf/skills/directory - Fix
AGENTS.mdbeing ignored by Cascade in specific cases
MCP
- Add support for system-level Skill definitions via MDM-managed configs for Enterprise
- Improve context management for MCP servers
Stability & Performance
- Improve autocomplete error handling and performance
- Improve SSH & Remote performance
Bug Fixes and Improvements
- Fix extension installation version selection
New Model Picker
We introduced a new model picker that groups models by family and adds a hovercard with toggles for specific variants, like reasoning effort and speed. Separately, we also added the ability to pin models.Cascade Improvements
- Added
POST_CASCADE_RESPONSE_WITH_TRANSCRIPTcascade hook - Added Cascade hooks configuration visibility on team settings page
- Reduced the priority of Git commits in @ mention search
- Added a
windsurf.cascade.readClaudeCodeConfigflag to disable reading Claude configuration
MCP Improvements
- Added an MCP Refresh button
- Auto-trigger OAuth login when adding HTTP/SSE MCP servers
- Fix bugs with parsing on Windows and startup
Platform Improvements
- Merged changes from VS Code 1.108
- Improved startup reliability for Cascade
- Fixed Windows update initialization path that could block updates
- Released binaries for Linux ARM64
A release was made.
Bug Fixes and Improvements
- Fix compatibility with GitHub Pull Requests extension
Bug Fixes and Improvements
- Fix for self-updating on Windows
- Fix for macOS UI flickering
Cascade Improvements
- Plan Mode now supports automatic switching back to Code Mode when you start implementing a plan
- Added support for reading skills from the
.agents/skillsdirectory - Tracking triggered rules in the
post_cascade_responsehook via a newrules_appliedfield - Diff zones will now automatically close on commit
Linux ARM64 Support
- Full Linux ARM64 client support with deb and rpm packaging
Enterprise & Team Improvements
- Cloud configuration for Cascade Hooks is now available for enterprise teams via the cloud dashboard
- Support for Devin service key authentication
Bug Fixes and Improvements
- Fixes for
post_write_codehooks to handle all code editing tool formats - Fixes and improvements for MCP server resource loading
- Fixed osascript privilege escalation being incorrectly triggered on Linux for shell command installation
- Addressed multiple memory leaks
- Improved RTL language rendering in todo lists
A release was made.
New Models
- GPT-5.3-Codex-Spark is now available in Arena Mode’s Fast Arena and Hybrid Arena battle groups
Claude Opus 4.6 (fast mode)
Claude Opus 4.6 (fast mode) is now available in Windsurf in research preview with limited-time promotional pricing for self serve users until Feb 16:- No thinking: 10x credits
- With thinking: 12x credits
Bug Fixes and Improvements
- Fix issues with Arena Mode Battle Groups
Bug Fixes and Improvements
- Improve UI styling for announcement popups and notifications
- Close model picker when selecting a battle group
Bug Fixes and Improvements
- Various bug fixes and performance improvements
Wave 14: Arena Mode
Arena Mode brings side-by-side model comparison directly into your IDE, plus Plan Mode for smarter task planning.Arena Mode
Run two Cascade agents side-by-side with hidden model identities and vote on which performs better. Arena Mode lets you discover which models actually work best for your workflow, codebase, and tasks—not just what benchmarks or influencers say.- Battle Groups: Choose specific models to compare or let Windsurf randomly select from curated groups like “fast models” vs “smart models”
- Personal & Global Leaderboards: Your votes contribute to both a personal leaderboard (your preferences) and a global one (across all Windsurf users)
- Sync or Branch: Send followup prompts to both agents simultaneously, or branch and explore different paths individually
Plan Mode
Plan Mode is a new Cascade mode alongside Code and Ask. Use it to create detailed implementation plans before diving into code.Pro tip: Typemegaplan in the Cascade input box to trigger an advanced form that asks clarifying questions to create a more aligned, comprehensive plan.Bug Fixes and Improvements
- Admins can now set a default model that applies to all team members when they first open Windsurf
- Various bug fixes and performance improvements
A release was made.
Bug Fixes and Improvements
- Fix bug with permanently disconnected cascades
Enterprise Features
- Enterprise admins can now specify organization-wide allow and deny lists for command auto-execution. Learn more
Bug Fixes and Improvements
- Bug fixes and performance improvements for diff zones
- Improved overall stability and reliability
- Added
post_setup_worktreehook for initializing worktrees in Cascade
Bug Fixes and Improvements
- Improvements to GPT-5.2-Codex harness
- Admins can now manage Windsurf restrictions via Windows Group Policy
GPT-5.2-Codex
Adds support for GPT-5.2-Codex with four reasoning efforts (low, medium, high, and xhigh). GPT-5.2-Codex is OpenAI’s latest model designed for agentic coding. It excels at working in large codebases over long sessions. For most tasks, we recommend using the medium reasoning effort.Bug Fixes and Improvements
- Various bug fixes and performance improvements
- Improved stability and reliability
Bug Fixes and Improvements
- Various bug fixes and performance improvements
- Improved stability and reliability
Gemini 3 Flash
Gemini 3 Flash is now available for all users. This model combines Gemini 3 Pro-grade reasoning with Flash-level speed and efficiency, making it ideal for agentic workflows and coding tasks.- Blazing Fast Responses: Experience near-instant feedback with 3x faster performance than previous generations, perfect for iterative development compared to Gemini 3 Pro
- Superior Coding Intelligence: Outperforms even Pro-tier models on key coding benchmarks (78% on SWE-bench Verified), providing more accurate code generation and debugging
- Deep Multimodal Understanding: Easily process complex video, data extraction, and visual Q&A tasks with frontier-level reasoning
Bug Fixes and Improvements
- Various bug fixes and performance improvements
- Tab fixes and improvements
Wave 13: Merry Shipmas
Wave 13 brings first-class support for parallel, multi-agent sessions in Windsurf, along with Git worktrees, side-by-side Cascade panes, and a dedicated terminal profile for more reliable agent execution.SWE-1.5 Free
Our near-frontier model, SWE-1.5, is now available for free to all users for the next 3 months. SWE-1.5 Free has the full intelligence of SWE-1.5, with the same coding performance on SWE-Bench-Pro, but delivered at standard throughput speeds. The original variant of SWE-1.5 hosted on Cerebras will continue to be available for paid users. SWE-1.5 Free will replace SWE-1 as the default model in Windsurf starting today.Git Worktree Support
Windsurf now supports Git worktrees, letting you spawn multiple Cascade sessions in the same repository without conflicts. Git worktrees check out different branches into separate directories while sharing the same Git history.Multi-Cascade Panes & Tabs
You can already run multiple Cascade sessions in Windsurf at the same time. Now, you can view and interact with them in separate panes and tabs within the same window. This lets you monitor progress and compare outputs of sessions side-by-side, or even turn Windsurf into a big Cascade dashboard.Cascade Dedicated Terminal (Beta)
Windsurf introduces a new approach for letting agents execute terminal commands. Instead of your default shell, Cascade will now run commands in a dedicated zsh shell specifically configured for reliability. The Cascade Dedicated Terminal can use the environment variables you set in your .zshrc configuration and is interactive, which means you can answer any prompts from shell scripts without having to break your flow. This should improve the reliability and speed of shell commands, especially for users with complicated prompts (e.g., powerlevel10k).In this version, the Cascade Dedicated Terminal will be opt-in for Windsurf Stable on macOS. If you are experiencing issues with the older legacy terminal, we recommend switching to the new terminal early. We expect to make this feature the default in the future, while maintaining the legacy terminal extension for backwards compatibility purposes. You can opt-in in the Windsurf User Settings -> Disable Windsurf Legacy Terminal Profile.Context Window Indicator
When a model’s context window grows too long, earlier context can be dropped without warning and performance can degrade. Cascade already extends the window by occasionally summarizing messages and clearing history. This release adds a visual indicator to see how much of your context window is currently in use, helping you anticipate limits and decide when to start a new session.Cascade Hooks
Execute custom commands at key points during Cascade’s workflow, including on model response for auditing purposes.System-level Rules & Workflows
Enterprises can deploy rules and workflows via MDM policies, allowing organizations to place rules and workflows files on users’ machines.Bug Fixes and Improvements
- Enhanced diff zone behavior with configurable scroll-to-next-hunk settings (default off)
- Preserved colors and styling in Cascade terminal output
- Multiple fixes to the Model Context Protocol implementation
- Supports lowering permissions for Cascade’s Web Fetch tool
- Fix race condition in the dedicated terminal implementation
- Support force killing commands in the dedicated terminal
- Improved markdown completion
- Fix opening old Cascade diffs
Bug Fixes and Performance Improvements
- Fix race condition in the new dedicated terminal implementation
- MCP fixes
- Added setting to control scroll-to-next-hunk behavior in diff zones (default off)
- Improve markdown completion
- Preserve terminal colors and styling in the Cascade terminal output pane
- Diff zone fixes and improvements
- Support turbo mode for web fetch requests made by Cascade’s Web Fetch tool
- Support force killing commands in the new dedicated terminal
New Features
Windsurf Dedicated Terminal
- Introduces a new terminal for Cascade that improves command execution reliability (OS X only)
Multi Cascade Panels and Tabs
- Adds support for multiple Cascade panels and tabs, allowing users to work with multiple Cascade sessions simultaneously
Cascade Hooks on Cascade Response
- Allows users to configure Cascade Hooks on model response, specifically for logging all Cascade responses for auditing purposes
System-level Rules + Workflows
- Allows enterprises to place rules and workflows files on users’ machines with MDM policies
Bug Fixes
- Fix opening old Cascade diffs
- Fix various stability issues with the new terminal implementation
Features
Added a new “Promo” label to LLM models that are newly available or have special discount pricingPatch Fixes and Improvements
- Enhances code block file path display to hide line numbers for whole files
- Improves citation and language parsing in code blocks with a more robust regex pattern
- Updates the UI for code block title bars to properly handle long paths with truncation
- Fixes fallback diff handling for nonexistent files in code actions
- Improves the auto-run command menu interface and its display logic
Patch Fixes and Improvements
Agents & Tool Execution
- Fixed Command-I functionality
- Fixed Ctrl+C during tool execution not working properly
- Fixed Go (fallback) processes not being killed properly
- Fixed handling of parallel tool call errors
UI & Rendering
- Fixed incorrect indentation from code blocks in terminal rendering
- Fixed nested lists not rendering on new line in terminal markdown
- Fixed content spacing issues
- Fixed streaming flashes
Messaging & Platform
- Fixed rate limit error message to say “no credits were used” instead of “credits have been refunded”
- Fixed continuously rechecking for updates on macOS
- Added a user-facing message when API providers are exhausted
Tooling & Onboarding
- Improved MCP tool call visibility (show tool name, args, etc)
- Added loading indicators when thinking or during long running operations
- Allowed clicking items in the Windsurf onboarding pane
- Respected gitignore patterns in the workspace directory tree
Patch Fixes and Improvements
- Reduce occurrence of “prompt is too long” errors
- Request all supported scopes if no scopes are provided in MCP OAuth config
GPT-5.2
GPT-5.2 is now available in Windsurf. This model will be available for 0x credits in Windsurf (to paid users) for a limited time.GPT-5.2 represents the biggest leap for GPT models in agentic coding since GPT-5 and is a SOTA coding model in its price range. The version bump undersells the jump in intelligence. We`re excited to make it the default across Windsurf and several core Devin workloads. - Jeff Wang, CEO of WindsurfDownload the latest version on Windsurf to try it out!Bug fixes and improvements
- General Windsurf stability and performance improvements
- General Tab (Supercomplete) improvements and stability
- Fixes issues with Cascade running commands that could not be cancelled during certain long-running processes
Diff Zones
- Fixes issues with diff zones not rendering correctly or jumping to the end of a file when editing
Lifeguard
- Fixes various Lifeguard bugs and stability issues
Tab (Supercomplete)
- Improves reliability of Tab autocomplete
- Makes Tab more responsive and faster in the appropiate circumstances
MCP Servers
- Added support for MCP prompts
- Added toggles to enable/disable MCPs in the Cascade header
Bug fixes and improvements
- General stability and performance improvements
- Fixes issues with login timing out too quickly during onboarding
Features & Tools
Cascade Hooks on User Prompts
- Users can now configure Cascade Hooks on user prompts for logging all user prompts and blocking policy-violating prompts.
Worktree / Arena Mode
- Added support for Worktree and Arena Mode.
MCP Servers
- Added support for GitLab remote MCP.
- Added OAuth support for GitHub remote MCP.
- Fixed an issue where every MCP would reauth on opening Windsurf.
General Improvements
- General performance and stability improvements.
- Improvements for Tab (Supercomplete) autocomplete reliability.
GPT-5.1-Codex Max
Introducing GPT-5.1-Codex Max in three reasoning tiers (Low, Medium, High). Low variant available at no cost to paid users for a limited time.Claude Opus 4.5
You can now use Claude Opus 4.5 in Windsurf!Opus 4.5 is the most capable model in Windsurf yet and is now available at Sonnet pricing for a limited time (2x credits compared to 20x for Opus 4.1).This model is available to all paid Windsurf subscribers.AI Models
Gemini 3 Pro
- You can now use Gemini 3 Pro (Low and High) in Windsurf! This is a preview release that is available to paid Trial, Pro, and Teams subscribers and will soon be extended to Enterprise users as well.
- Resolved an issue where Gemini 3 Pro caused internal Cascade errors for some users.
SWE-1.5
- Fixed notebook tool functionality for SWE-1.5.
- Addressed an issue where SWE-1.5 would unexpectedly stop or return “No response requested”.
Sonnet 4.5
- Added support for Sonnet 4.5 with a 1M token context window.
- Reduced the frequency of unnecessary Markdown file creation by Sonnet 4.5.
GPT-5.1 Codex and Codex Mini
- Added support for GPT-5.1 Codex and Codex Mini with low reasoning effort configuration.
Features & Tools
Codemaps
- Improved reliability of saving and retrieving Codemaps.
- Fixed Codemap sorting and increased the limit of visible open Codemaps.
- Resolved issues with mentioning Codemaps in Cascade.
MCP Servers
- Fixed scope handling and OAuth authentication flows for various MCP servers.
- Resolved issues preventing installation of new MCP servers.
- Added support for handling embedded resource content in tool call responses.
Tab Completion
- Improved latency, responsiveness, and accuracy of autocomplete.
Vibe and Replace
- Fixed reliability issues with Vibe and Replace.
Fast Context
- Added support for
.codeiumignoreand.gitignorefor Fast Context.
General Improvements
- Fixed various UI alignment issues with icons and styles.
- General performance and stability improvements.
- Fixed issues with the file search tool.
Worktree Preview
You can now use Windsurf’s AI agent with worktrees to explore and modify code across multiple branches simultaneously.MCP Server Improvements
- Fixes issues with scopes and OAuth authentication flows for multiple MCP servers
Tab Completion
- Improved latency, responsiveness, and accuracy of autocomplete.
Context Window Indicator Beta
- Context window indicator now shows the current context window used by the Cascade AI agent.
GPT-5.1 Priority Mode
- Added priority processing support for GPT-5.1 models, providing guaranteed low-latency responses for faster (~50 tokens/sec), more reliable AI assistance.
- Priority processing costs 2x the standard rate, which will be reflected in the Windsurf credit system.
Patch Fixes and Improvements
- Fixed tool call calling issues for GPT-5.1-Codex and GPT-5.1-Codex Mini models
GPT-5.1 and GPT-5.1-Codex
GPT-5.1 and GPT-5.1-Codex are now available in Windsurf. GPT-5.1 will become the default model in Windsurf for one week, and paid users get free access during this period.GPT-5.1 and GPT-5.1-Codex deliver a solid upgrade from GPT-5 for agentic coding workflows. They’re noticeably better at understanding what you’re asking for and working with you to get it done. The new variable thinking feature dynamically adjusts reasoning depth—providing quick responses for simple tasks and more thoughtful analysis when complexity demands it.Patch Fixes and Improvements
Removed Features
- Knowledge Base: Removed the Knowledge Base feature
Bug Fixes
- PowerShell + Turbo Mode: Fixed an issue where PowerShell was not running commands when Turbo Mode is enabled
Shortcuts
- Attach current file to Cascade: New shortcut
Option/Alt+Cmd+Lwhen in an editor to attach the current file to Cascade
Features
MCP Improvements
- Loading state indicators: Show loading state per installed MCP to improve visibility during initialization
- Refresh only edited MCPs: When
mcp_config.jsonis modified, only the affected MCP server instance is initialized/refreshed - no other instances are refreshed - Increased initialization timeout: MCP initialization timeout increased to 60s
- Refresh button for error states: Show refresh button for MCPs in error state to allow manual recovery
Cascade Hooks
- Enterprise feature: New Cascade Hooks feature available
- Public documentation: Draft documentation with details and descriptions of hooks available in Windsurf Docs
Bug Fixes
- Vim extension typing lag: Fixed bug causing typing lag when Vim extension is enabled
SWE 1.5 Image Support
- Image understanding: Images are now supported in SWE 1.5, enabling visual content analysis
Features
Expanded Codemaps
Codemaps now include powerful new capabilities:- Chat with map - Interact directly with your codebase visualizations
- Mermaid diagrams - Generate visual diagrams within maps for better code understanding
- Cascade suggestions - Get AI-powered suggestions directly in your maps
- Map option in chat/edit nudges - Easily create maps from chat and edit interactions
- Smart mode option - Enhanced intelligent assistance when working with maps
Cascade Summarization Fix
Improved Cascade summarization to better handle longer conversations. Previously, summaries could be too aggressive and drop important context. Now maintains better continuity across long sessions with multiple file changes and user messages.Sonnet 4.5 Support for SWE 1.5 Planning
You can now enable beta Sonnet 4.5 planning when using SWE 1.5. To enable, select SWE 1.5 first, then select the Sonnet 4.5 addon.MCP Enhancements
- Path component handling - Improved support for MCP URLs with path components (e.g., Smithery MCPs)
- OAuth flow improvements - Better OAuth flow for streamable HTTP MCPs (e.g., Canva, ServiceNow’s internal MCP)
Bug fixes and improvements
Performance Improvements
- Sticky scroll lag fixes - Resolved lag spikes when using sticky scroll with Vim bindings
- General slowness fixes - Addressed performance issues caused by VSCode OSS update
- Terminal rendering optimization - Fixed rendering loop that caused 500ms+ delays on first terminal open
Terminal Fixes
- PowerShell improvements - Fixed Windows terminal integration issues where commands would appear stuck
- Shell theme compatibility - Resolved edge cases with custom shell themes (zsh, fish, powerlevel10k, etc.) that could cause Windsurf to break or show stuck commands
Editor Stability
- Terminal freeze fix - Fixed an issue where the editor would freeze when opening the terminal
- CMD+J fix - Resolved layout thrashing issue when opening terminal pane with CMD+J
Features
New beta plan mode (type /plan in the Cascade chat input to activate it).Expanded Codemaps:- Chat with map
- Mermaid diagrams in maps
- Cascade suggestions for maps
- Chat / edit nudge includes “map option”
- Smart mode option
Bug fixes and improvements
- Fixes for lag spikes when using sticky scroll
- Fixes for general slowness caused by VSCode OSS update
- Powershell fixes for Windows terminal integration / commands appearing stuck.
- Fixes for certain edge cases around shells with custom themes.
- Fixed an issue where the editor would freeze when opening the terminal
Patch Fixes and Improvements
- Support and fixes for AGENTS.md
- Improvements and bug fixes for Codemaps.
- Improvements to Fast Context. Enterprises can opt in using the Windsurf Team Settings. Users can toggle Fast Context automatically using “CMD/Ctrl + Enter” on the first message in a chat.
- New auto-linting behavior that speeds up Cascade.
- Fix for MCP Marketplace not respecting team whitelist options.
- Fixes for Jupyter Notebook tool.
- Fixes for Memories, Rules, and Workflows.
- General bug fixes and improvements.
- Performance optimizations and stability enhancements.
Dependencies
- Updated Code OSS to version 1.105.0 (Electron: 37.6.0, Chromium: 138.0.7204.251)
Patch Fixes
- Early preview of Windsurf’s new Tab model.
- Bug fixes and improvements.
Patch Fixes
- Fixes issues with the experimental Cascade tool for finding context not working properly.
- Improvements and bug fixes for the beta Codemaps feature.
- Improvements to the Lifeguard (Beta) feature.
- Fixes issue with custom MCP servers not being displayed correctly in the new MCP panel.
- Fixes issue where some bash commands would get stuck.
- Fixes issue where certain models couldn’t create or edit Jupyter notebooks.
- General bug fixes and improvements.
Patch Fixes
- Various improvements to the experimental Cascade tool which searches and finds relevant files.
Experimental Cascade Tool
- Experimental Cascade tool to quickly search for files and find relevant context.
Lifeguard
- Beta preview of lifeguard, a tool to help you find and resolve bugs inside your IDE.
Codemaps
- Beta preview of codemaps: open the codemaps pane to try it out!
Patch Fixes
- Fix using MCP tools with certain models.
- Fixes to terminal issues on Windows.
GPT-5-Codex is now in Windsurf!
GPT-5-Codex is now available for free (0x credits) for a limited time for paid users!Free users can use GPT-5-Codex as well for 0.5x credits.Speak to Cascade
Voice
- Users can now speak into the chat rather than having to type things out.
@-mentioning conversations
- @-mention the first conversation so Cascade has full context of it as it goes to write tests for you.
Deeper Browser integration
- Chat with Cascade about tabs that are open in the Browser using @-mentions
JetBrains improvements
- Planning Mode, Workflows, and file-based Rules are now available for Cascade on JetBrains
Improvements
- Now you can @-mention terminal in Cascade.
- You can turn on the Auto-Continue setting to have Cascade automatically continue its response if it hits a limit.
- Support for more MCP servers with easier and more secure authentication by integrating the new Streamable HTTP transport (replaces SSE) and MCP authentication (replaces access tokens or API keys in the config).
- Important for enterprise customers who use Windsurf across lots of repos. Now, you can enforce ignore rules across all repositories by placing .codeiumignore in the ~/.codeium/ folder
Updates
- Removal of legacy mode
- Icons in @-mentions
- Theme-aware codeblocks with refreshed design
- Native terminal in Cascade panel
BYOK (Anthropic Key)
- Based on the latest release of the Windsurf Editor, v1.9.2
New Family of SWE-1 Models
- Based on the latest release of the Windsurf Editor, v1.9.0
Improvements
Some improvements to app icons, fixes to SSH server connection, and UI Polish for Cascade Plugins.New Features
Cascade Plugin Panel
- New panel in Cascade for managing MCP Servers
- Easier one-click uninstall and install
- Easier search
- MCP now has MCP resources and multimodel responses
- More MCP Server options coming soon
Cascade Customization Panel
- Revamped Memories panel to include rules, workflows, and memories
Revamped rules
- Support for workspace-style rules in
.windsurf/rules - Custom triggers when to activate a rule
Workflows
- Support for quick instructions for Cascade to take on mini-tasks
.windsurf/workflowshouses workflows that Cascade can do in the current workspace- Workflows can be found as slash commands (/) in the Cascade input
Cascade Improvements
- Conversation sharing with a teams-only accessible URL
Misc
- Redesigned Model Selector
- Continue button when reaching individual tool call limit
- Hunk accept/reject widget now has a compact mode to cover less code
- Tab respects files mentioned in
codeiumignore
New App Icon & Upgraded Free Tier
Based on the latest release of the Windsurf Editor, v1.7.2Experimental Improvements
- Bug fixes
- Performance optimizations
- VS Code 1.98.0 Updates
Updated and Simplified Pricing
- We’re simplifying our pricing model by removing Flow Action Credits
- Change takes effect April 21st, 2025
- Plans now come with prompt credits with add-on credits available for purchase
User Prompt Credits
- Plans now come with prompt credits, which are consumed per every message sent and not via every tool call
- Add-on credits are available for purchase
- Auto-top off (with max limits) can be enabled via profile
Existing Plans
- Existing plans are migrating over to the new pricing model
- For more information, please visit the Pricing page

