- Add snapshot file for native tools returned by system prompt getter function
- Remove inline comment from test:unit npm script that was breaking the command "npm run test:unit -- --update-snapshots"
- GLM-4.6
- kat-coder-pro
- Add parsing of env variable patterns to the mcpconfig.json
- TLS Proxy support issues for VSCode
- Add supportsReasoning flag to OpenAI reasoning models
- Fix thinking not available for some models in the OpenAI provider
- Fix invalid signature field issues when switching between Gemini and Anthropic providers
- Extract OpenRouter model filtering into reusable utility and use it in different model pickers
- Fix a11y for auto approve checkbox
- Improve ModelPickerModal provider list layout
- Migrate WhatsNewModal to new shared dialogue component
Co-authored-by: Arafatkatze <arafat.da.khan@gmail.com>
- Expose `getAvailableSlashCommands` rpc endpoint to UI clients
- Made slash command menu and context menu accessible and screenreader-friendly
- Made expanding/collapsing UI components accessible
- Model identity and routing for devstral-2512 free model
- Extension pricing/UI bug where extension incorrectly shows zero price for devstral-2512
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
- OpenAI GPT-5.2
- Devstral-2 `devstral-2512` (formerly stealth model "Microwave")
- Improvements to chat modal model picker
- Amazon Nova 2 Lite
- DeepSeek 3.2 to native tool calling allow list
- Responses API support for Codex models in OpenAI provider (requires native tool calling)
- Xmas Special Santa Cline
- Welcome screen UI enhancements
- Initial checkpoint commit now non-blocking for improved responsiveness in large repositories
- Gemini Vertex models erroring when thinking parameters are not supported
- Restrictive file permissions for secrets.json
- Ollama streaming requests not aborting when task is cancelled
- OpenAI provider to centralize temperature configuration and include missing GPT-5 model settings
- OpenAI native handler to use metadata for model capabilities
- Vertex provider to use metadata for model capabilities
Co-authored-by: Arafatkatze <arafat.da.khan@gmail.com>
* v3.39.0 Release Notes
* v3.39.0 Release Notes
* feat: enhance Announcement component with new microwave model and account login functionality
- Updated Announcement component to include a new free microwave model button
- Adjusted active tab logic in OpenRouterModelPicker to default to "free" if a free model is selected
* Add demo link
---------
Co-authored-by: Saoud Rizwan <7799382+saoudrizwan@users.noreply.github.com>
- Add cline-bot icon font assets (SVG, TTF, WOFF) generated from IcoMoon
- Register custom icon font in VS Code extension manifest
- Replace PNG-based command icon with font-based cline-icon
- Update terminal icon references from generic "robot" to "cline-icon"
This provides a consistent branded icon across the extension and improves visual identity by using the official Cline bot logo instead of the generic robot icon from codicon that was updated by VS Code.
* chore: clear onboarding models on deactivate
* feat(storybook): add OnboardingView story
- Added OnboardingView component to Storybook with new story
- Integrated onboarding models from shared constants
- Updated MockApp to conditionally render OnboardingView based on onboardingModels state
- Renamed WelcomeScreen story to Welcome for clarity
- Added interaction tests for onboarding buttons (Get Started/Use your own API key)
- Configured onboarding models in mock state to support new story
This enables visual testing and documentation of the user onboarding flow within Storybook.
* Update Storybook missing vscode theme color
* Update task name
* typo
* feat: add support for Responses API for openai-native provider [ENG-1227]
- Upgrade openai dependency to v6.9.0 to use the Responses API
- Implement internal handling for reasoning/thinking and redacted output
- Align Anthropic handler message types with the latest SDK interfaces
- Clean up obsolete tooling imports related to tool-use handling
- Enable newer OpenAI capabilities while keeping provider APIs consistent
* Add openai_native_response_api feature flag
* clean up
* clean up 2
* add back gpt-5.1 models
* use call_id
* v3.38.0 Release Notes
- Gemini 3 Pro Preview model
- AquaVoice Avalon model for voice-to-text dictation
- Automatic context truncation when AWS Bedrock token usage rate limits are exceeded
- SAP AI SDK JS packages upgraded to latest major version
- SAP provider OrchestrationClient now matches OrchestrationModuleConfig type and no longer uses invalid promptTemplating property
- Removed new_task tool from system prompts, updated slash command prompts, and added helper function for native tool calling validation
* Update CHANGELOG.md
---------
Co-authored-by: Arafatkatze <arafat.da.khan@gmail.com>
* feat: add thought signature support for Gemini SDK
Update @google/genai dependency from v1.15.0 to v1.30.0, including nested deps like google-auth-library. Enhance API interfaces with JSDoc comments and new fields such as signature, id, and redacted_data in ApiStreamThinkingChunk to support thought signatures from Gemini SDK as requested. This improves integration with Gemini's reasoning capabilities and ensures compatibility with updated SDK features.
* add changeset
* meaning val check
* typo
* either
* Do not use think budget with gemini-3
* feat: upgrade sap-ai-sdk major version
* feat: upgrade sap-ai-sdk major version
* feat: upgrade sap-ai-sdk major version
* feat: upgrade sap-ai-sdk major version
* feat: upgrade sap-ai-sdk major version
* feat: upgrade sap-ai-sdk major version
* feat: upgrade sap-ai-sdk major version
* feat: upgrade sap-ai-sdk major version
* feat: upgrade sap-ai-sdk major version
* feat: upgrade sap-ai-sdk major version
* feat: upgrade sap-ai-sdk major version
* v3.37.0 Release Notes
- GPT-5.1 with model-specific prompting: tailored system prompts, tool usage, focus chain, and deep-planning optimizations
- Nous Research provider with Hermes 4 model family and custom system prompts
- Model-family breakouts for deep-planning prompting, laying groundwork for enhanced slash commands
- Expanded HTTP proxy support throughout the codebase
- Improved focus chain prompting for frontier models (Anthropic, OpenAI, Gemini, xAI)
- Duplicate tool results prevention through existence checking
- XML entity escaping in model content processor
- Commit message generation in command palette
- OpenAI Compatible provider temperature parameter type conversion
- Added missing proto generation step in CONTRIBUTING.md
- New `npm run dev` script for streamlined terminal workflow (fixes#7335)
* Adding image optimizations
* Adding image optimizations
---------
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: Arafatkatze <arafat.da.khan@gmail.com>
- Add step 4 in Local Development Instructions to run npm run protos
- Add proto generation step in Extension section
- Create new npm run dev script that runs protos + watch
- Update Terminal Workflow documentation to mention npm run dev
Fixes#7335
* fix: function declarations tool converter for native tool calling
- Replace crypto-based ID generation with nanoid in McpHub that is causing invalid function name issue with Gemini models
- Refactor Google tool parameter schema mapping with typed constant
- Improve parameter properties handling in toolSpecFunctionDeclarations
- Add support for nested properties and enum values
- Skip $schema property during parameter processing
This change fix type safety for Google Gemini function calling parameter definitions.
* disable native tool calling by default
* Always starts with a letter
* whoops
* reuse id
* fix run time package issue
- Add: Hooks allow you to inject custom logic into Cline's workflow
- Add: new provider AIhubmix
- Add: Use http_proxy, https_proxy and no_proxy in JetBrains
- Fix: Oca Token Refresh logic
- Fix: issues where assistant message with empty content is added to conversation history
- Fix: bug where the checkbox shows in the model selector dropdown
- Fix: Switch from defaultUserAgentProvider to customUserAgent for Bedrock
- Fix: support for `<think>` tags for better compatibility with open-source models, and refinements to the GLM-4.6 system prompt
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Better documentation in shared/net
Better docs, wire up when running standalone.
Use esbuild to statically decide which proxying we are doing.
Add the changeset.
Add the ability to mock this fetch.
* feat(hooks): Initial implementation of hooks UI using background terminal UI
* Hooks UI improvements
* Hook discovery improvements
* Separate hooks UI into separate files
* Fix failing tests
* Add missing docs
* Hooks hardening and handling edge cases
* Improvements to hooks error messaging
* Improve TaskCancel hook UI
* Simplify and improve console.log() output from hooks
* Changes from usability feedback
* test: verify linting fixes
* fix: resolve cancel functionality in PostToolUse hook using flag pattern
* fix: add cancel button support for TaskResume and UserPromptSubmit hooks
* Getting things working again
* Fixing cancel behaviors
* Trying another fix
* Demoed
* Add examples to .clinerules/hooks/ directory
* Refactored away loadTaskStateWithoutWorkflow
* Fix unsafe return in finally block
Refactored PostToolUse hook execution to avoid using return statement
in finally block, which the linter flags as unsafe. Instead of early
return, wrapped the hook logic in a conditional check for abort status.
This resolves the biome lint error:
lint/correctness/noUnsafeFinally - Unsafe usage of 'return' in finally block
* Remove unnecessary files from branch
* remove cancelHookExecution() in favor of cancelTask()
* Remove unnecessary polling pattern for hook cancellation
* Fixing cancel behavior
* Fixing cancel behavior
* Fix resume
* Sqaush commits to improve merge tool behavior
Plan out the work needed to resolve race conditions
Fix race conditions (1)
Fix race conditions (2)
Fix race conditions (3)
Fix race conditions (4)
Fix race conditions (5)
Fix unit tests
Updated implementation plan doc
Code quality improvement
Temporary logging to troubleshoot race conditions in hooks UI blocks
Fix streaming output for hooks
Remove in-progress planning file
Revert breakage
Remove unneeded arg output in PreToolUse hook UI
Fix race condition with PreToolUse and PostToolUse hooks
Fix PreToolUse attempt_completion use case
Fix ChatTextArea prompt input
Fixed part of the rersume bug
Clean-up before code review
Change cancel button to abort button
Fixed TaskCancel behavior
Fixing resume behavior (partially fixed)
Make notch arrow visible in hooks block up expand/contract toggle
Prevent cancel/abort from clearing unsent user message
Change 'Cancelled' to 'Aborted' in hook UI block
Use color styles consistent with background terminal
Fix npm compile issue
Remove unneeded hot-cold tracking from hooks
Remove unused constant as per PR feedback
Minor changes
Skip combineHookSequences() if hooks feature setting not enabled
Improve mutex pattern for general use and fix additional race condition
Refactor to reduce duplication (keep it DRY)
Refactor to reduce duplication (keep it DRY) part 2
Fix tests
Skip some tests on Windows (not yet supported)
Hooks unsupported on Windows
Fix broken unit tests
* Fix broken unit tests
* Bee and Eve fixing 'API Request...' test in e2e tests
* Add comment as per ellipsis-dev PR feedback
* Changes as per PR feedback from Evan
* Remove if that always resolves to true
* Don't use vscode API; use cross-compatible solutions instead
* Remove mention of obsolete clineAsk handler
- Add native tool calling support with configurable setting.
- Auto-approve is now always-on with a redesigned expanding menu. Settings simplified and notifications moved to General Settings.
- added zai-glm-4.6 as a Cerebras model
- Created GPT5 family specific system prompt template
- Fix: show reasoning budget slider to models with valid thinking config
- Requesty base URL, and API key fixes
- Delete all Auth Tokens when logging out
- Support for <think> tags for models that prefer that over <thinking>
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
* (WIP)feat(api): add tool calling support to openai API handlers
Add ChatCompletionTool parameter to createMessage interface and implement tool calling functionality across API providers. Includes new StreamingToolCallHandler for processing tool call deltas and utility functions for converting tool specifications to OpenAI format.
* Replace path with absolutePath
* Simplify tool spec
* Parse stream and use variant matcher
* Simplify StreamingToolCallHandler
* Add CLINE_NEXT_GEN
* Enable parallel_tool_calls
* ApplyPatchHandler
* Simplify ApplyPatchHandler
* store tool use chunk
* feat(anthropic): add streaming tool use support with OpenAI-compatible format
Add comprehensive support for streaming tool use blocks from Anthropic API:
- Add tool_use case handler in AnthropicHandler to convert streaming tool
calls to OpenAI-compatible format during content block start events
- Implement ToolUseHandler class to accumulate and parse streaming tool use
deltas with incremental JSON parsing support
- Add proper handling for undefined tool message content in openai-format
converter to prevent errors
- Integrate ToolUseHandler into TaskState for managing streaming tool calls
across the request lifecycle
This enables real-time processing of tool calls as they stream from Anthropic's
API while maintaining compatibility with OpenAI's tool calling format.
* feat(anthropic): add tool support with OpenAI tool conversion
Add support for passing tools to Anthropic API with automatic conversion from OpenAI tool format. Implements tool_choice configuration to force tool usage when tools are provided.
Changes:
- Add tools parameter to createMessage method in AnthropicHandler
- Implement openAIToolToAnthropic converter to transform OpenAI ChatCompletionTool format to Anthropic Tool format
- Set tool_choice to "any" when tools are provided to ensure Claude uses one of the available tools
- Fix tool result handling to avoid invalid tool_use_id "cline" by converting to text blocks for backward compatibility
- Add documentation for tool_choice options (none, auto, any)
This enables proper tool calling functionality with Anthropic models while maintaining compatibility with the existing OpenAI tool definitions.
* oops
* fix: remove .mjs extensions from import paths
Remove `.mjs` file extensions from OpenAI and Anthropic SDK import paths to ensure compatibility with current module resolution. This updates imports across multiple files:
- openai/resources/chat/completions.mjs → openai/resources/chat/completions
- @anthropic-ai/sdk/resources/index.mjs → @anthropic-ai/sdk/resources/index
* revert styles.css change from merge main
* Update config and fix tests
* gpt-5 variant for cline provider only
* Fix GPT-5 web fetch tool
* Add native next gen for supported openai providers
* add native tool calling support to more openai comp providers
* list supported providers for gpt-5
* clean up
* Update ApplyPatchHandler
* clean up
* Update ask_followup_question
* Add system prompt test for cline claude 4.5 sonnet
* noToolsUsed
* Add support to gemini api provider
* add native tool calling to more providers
* Skip XML parsing
* clean up
* add back TASK_PROGRESS_PARAMETER to native tools
* feature flag NATIVE_TOOL_CALLS_NEXT_GEN_MODELS
* add allowNativeToolCalls to system prompt context
* add grok-code
* feat(context): ensure tool_result blocks follow tool_use blocks
Add validation to ensure every tool_use block in assistant messages has a corresponding tool_result in the following user message. When tool_result blocks are missing, they are automatically added with "result missing" content to maintain proper message pairing required by the Anthropic API.
This prevents API errors caused by orphaned tool_use blocks and ensures conversation history integrity after context management operations.
* feat(task): refactor ApplyPatchHandler to use FileEditProvider with revert support
Refactor the ApplyPatchHandler to use FileEditProvider and DiffViewProvider
for file operations instead of direct filesystem access. This change adds:
- State tracking for applied commits to enable rollback functionality
- Automatic revert of changes when user denies approval or errors occur
- Enhanced response format that includes full file contents after patching
- Consistent file reading through DiffViewProvider for better editor integration
- Proper cleanup of applied state after successful operations
This improves the patch application workflow by ensuring changes can be
safely reverted and provides better visibility into the applied changes.
The refactoring also aligns file operations with the existing editor
integration patterns used elsewhere in the codebase.
* Finalizing...
* MCP tools converter
* remove ClineDefaultTool.MCP_USE
* Update snapshot
* comments
* use diffViewProvider
* Fix MCP tool use with images
* fix mcp tool converter with reserved keys
* Fix apply patch streaming issue
* add vercel-ai-gateway to supported provider & clean up
* simplfy apply patch handler
* Turn off thinking when native tools are enabled for anthropic provider
* clean up ToolUseHandler
* anthropic tool_choice logic
* improve tool result ordering
- Extract anthropic message mapping logic into a separate variable for better readability
- Add MessageParam import from Anthropic SDK
- Update ensureToolResultsFollowToolUse to maintain tool_result order matching tool_use blocks
- Improve documentation to clarify tool result ordering requirements
This refactoring improves code maintainability by separating the message transformation logic and ensures tool results are properly ordered to match their corresponding tool use blocks in the conversation flow.
<budget:token_budget>200000</budget:token_budget>
* refactor tool result validation for improved efficiency
Simplify ensureToolResultsFollowToolUse method by streamlining the logic
for validating and reordering tool_result blocks. Changes include:
- Use early returns to reduce nesting and improve readability
- Process tool results and other blocks in a single pass instead of multiple iterations
- Eliminate unnecessary deep cloning by mutating content array directly
- Add flag-based update detection to avoid unnecessary array rebuilding
- Simplify the reordering logic while maintaining the same validation rules
This refactoring maintains the same functional behavior (ensuring tool_result
blocks follow their corresponding tool_use blocks) while reducing complexity
and improving performance through more efficient array operations.
<budget:token_budget>200000</budget:token_budget>
* attempt_completion
* use helper function
* remove FreeModelIDs from merge conflicts
* clean up
* Add NATIVE_GPT_5 & exclude gpt-5-chat
* shouldAutoApproveToolWithPath for apply patch
* add openai-native to supported provider
* add lightweight incremental chunk extraction
* update apply patch handler
* revert version regex
* Fix browser action
* feat(proto-python): add script-generated Python gRPC stubs, Go-like client, docs, and PyPI publish workflow
- add scripts/build-python-proto.mjs
- invokes python -m grpc_tools.protoc over proto/**/*.proto
- outputs to src/generated/grpc-python
- mirrors Go layout under client/: connection.py, cline_client.py, services/_client.py
- supports PYTHON env override (use a venv interpreter easily)
- generates src/generated/grpc-python/pyproject.toml so output can be pip installed (pip install -e src/generated/grpc-python)
- package.json
- add protos-python script to run the generator
- docs
- add docs/exploring-clines-tools/python-protos.mdx with venv setup, generation steps, and import examples
- emphasize: everything in src/generated is produced by scripts (do not commit manual edits)
- CI: publish to PyPI only
- add .github/workflows/publish-grpc-python.yml
- workflow generates code via script, builds wheel/sdist from src/generated/grpc-python, and uploads to PyPI
- requires repo secret: PYPI_API_TOKEN (TWINE_USERNAME=__token__)
- optional version override input for workflow_dispatch
Notes:
- generation strictly produces all content under src/generated/grpc-python (including pyproject.toml)
- default package name in generated pyproject is cline-grpc-python (adjustable in the script if needed)
- recommended usage on macOS: PYTHON=/Users/nturumel/projects/oracle-github/cline/.venv-proto/bin/python npm run protos-python
* Delete .github/workflows/publish-grpc-python.yml
* Delete docs/exploring-clines-tools/python-protos.mdx
* Update tired-banks-show.md
---------
Co-authored-by: Andrei Eternal <206184+Garoth@users.noreply.github.com>
* Set up Tailwind v4
npx @tailwindcss/upgrade 1 ↵
≈ tailwindcss v4.1.13
│ Searching for CSS files in the current directory and its subdirectories…
│ Migrating stylesheets…
│ ↳ Migrated stylesheet: `./src/index.css`
│ Updating dependencies…
│ ↳ Updated package: `tailwindcss`
│ ↳ Updated package: `@tailwindcss/vite`
│ Migrating templates…
│ ↳ Migrated templates for: `./src/index.css`
│ Verify the changes and commit them to your repository.
* Migrate HeroUITooltip to radix-ui shadcn components
* import main.css
* Update e2e test text
* clean up
* Update mode switch test
* Fix auto approve modal z-index number
* Unify styles with theme
* fix spacing and sizes
* update logo id
* Fix e2e test
* Clean up
* npm install tailwindcss @tailwindcss/vite
* npx @tailwindcss/upgrade
≈ tailwindcss v4.1.14
│ ↳ Upgrading from Tailwind CSS `v4.1.14`
│ Searching for CSS files in the current directory and its subdirectories…
│ Migrating stylesheets…
│ ↳ Migrated stylesheet: `./webview-ui/src/index.css`
│ Updating dependencies…
│ ↳ Updated package: `tailwindcss`
│ ↳ Updated package: `@tailwindcss/vite`
│ Migrating templates…
│ ↳ Migrated templates for: `./webview-ui/src/index.css`
│ Verify the changes and commit them to your repository.
* clean up
* clean up
* Remove DRY code and update descriptionForeground class name
* Update TaskHeader classnames
* unify font size
* clean up
* update test with clear test id
* size
* Fix tooltip trigger in settings
* Apply feedback - hide arrow for autoapprove menu
* Align chat toolbox icon stylings
* update data-testid
* set
* CheckpointError
* feat: arrow alignment issues
* text-wrap tooltip
* feat: mcp tooltip arrow fix
---------
Co-authored-by: Jose R. Perez <trupix@gmail.com>
- Added support for MiniMax provider with MiniMax-M2 model
- Remove Cline/code-supernova-1-million model
- Changes to allow users to manually enter model names (eg. presets) when using OpenRouter
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
* fix(scripts): fix proto-lint script execution on Windows
On Windows, directly calling 'scripts/proto-lint.sh' fails because it's not recognized as an internal or external command. This change wraps the script in an npm run command to ensure cross-platform compatibility. Added a new 'lint:proto' script for better organization.
* Update lint:proto script path to use relative path
* bash
* v3.34.0 Release Notes
- Cline Teams is now free through 2025 for unlimited users. Includes Jetbrains, RBAC, centralized billing and more.
- Use the “exacto” versions of GLM-4.6, Kimi-K2, and Qwen3-Coder in the Cline provider for the best balance of cost, speed, accuracy and tool-calling.
* fix: Adding Fallbacks
* fix: Adding Fallbacks
---------
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: Arafatkatze <arafat.da.khan@gmail.com>
github-actions[bot]