Changelog

All notable changes to TalkCody are documented here.

[0.4.0] - 2026-02-09

Added

• Telegram Remote Control: Support Telegram remote control for desktop TalkCody, supporting text, voice, and image messages with streaming response support.
• Feishu (Lark) Remote Control: Support Feishu remote control for desktop TalkCody, supporting text, voice, and image messages with streaming response support.
• Telegram Remote Control Documentation
• Feishu Remote Control Documentation
• Voice Transcription: Added Groq Whisper voice transcription support, available for free.
• Model Support: Added OpenRouter Pony Alpha free model.
• Explore Agent: Open Explore Agent, suitable for quickly understanding code repositories.

Fixed

• Fixed custom Provider configuration and request issues.
• Fixed input Token status display issue.
• Fixed share page theme display issue.

[0.3.5] - 2026-02-05

Added

• Support for Claude Opus 4.6 model
• Support for GPT 5.3 Codex model (OpenAI OAuth only)
• Support for disabling Tracing

Improved

• Model Search Optimization: Model search now supports filtering by provider name.

Fixed

• Fix Issue #47: Fixed Windows workspace-root path handling bug.
• Fix Issue #46: Fixed OpenAI OAuth login flow issues.

[0.3.4] - 2026-02-04

Improved

• User Abort Handling: Optimized handling of user aborts before hooks are triggered.

Fixed

• Cross-platform Compatibility: Fixed cross-platform path separator compatibility issues in project service and multiple other locations.
• Settings Store Fix: Fixed issue where the current task was not reset when switching projects.
• Fix Issue #43 & #44: Fixed bug causing loss of custom models.
• Gemini & Google AI Fix: Fixed Gemini model schema and Google AI API 400 errors.
• Fix Issue #45: Fixed GitHub skill import bug.

Acknowledgments

• Thanks to @qingzhongli for fixing cross-platform path compatibility issues and settings store bug.

[0.3.3] - 2026-02-02

Added

• LLM Tracing: Added complete LLM tracing functionality, supporting recording and viewing detailed AI request processes, execution time, Token usage, and Tool call details for debugging and prompt optimization.
• Kimi Coding Plan Usage: Added support for displaying Kimi Coding Plan usage statistics.

Improved

• Rust Backend LLM Architecture Refactoring: Refactored the LLM processing logic on the Rust side, laying the foundation for supporting 10-20 tasks in parallel.

Fixed

• Fixed window state save and restore bugs, improving multi-window experience.
• Fixed Issue #36: Optimized UI rendering logic in specific scenarios.
• Fixed Issue #39: Improved Bash tool execution feedback for complex commands.
• Fixed Issue #40: Fixed infinite refresh bug on Agents page.
• Fixed Kimi Coding Plan bugs.

[0.3.2] - 2026-01-27

Added

• Ralph Loop (Experimental): New persistent execution mode supporting AI Agent autonomous iteration for complex tasks. Agents can continue learning across multiple iterations, self-correcting based on execution feedback until completion criteria are met
• Kimi 2.5 Model Support: Added support for Kimi 2.5 model

Improved

• Web Fetch Tool Optimization: Improved web content fetching capabilities
• Create Agent Agent Optimization: Improved custom Agent creation experience

Fixed

• Fixed tool UI registration related bugs
• Fixed custom tool UI display bugs
• Fixed recent project list bug
• Fixed LLM retry mechanism issues
• Fixed Whats New dialog video Linux platform compatibility issue

[0.3.1] - 2026-01-25

Added

• Hooks System: New event-driven Hooks mechanism that supports triggering custom scripts at key events (user submission, before/after tool calls, session start/end, stop, etc.), enabling operation interception, parameter modification, audit logs, and other functionalities
• Auto Code Review: Support for automatic code review (Auto Code Review), which automatically runs the Code Review Agent after file changes
• LSP Tools: New complete LSP (Language Server Protocol) tools, supporting go-to-definition, find references, hover hints, document symbols, workspace symbols, go-to-implementation, call hierarchy analysis, and other operations
• API Usage Dashboard: New API usage visualization page, displaying chart statistics for Token usage, costs, and request counts, with filtering by today/this week/this month
• Custom Plan and Code Review Models: Support for configuring dedicated models for Plan Agent and Code Review Agent in settings
• Direct Plan Content Display: Display Plan file content directly in call-agent-tool-result without manually opening the file
• Custom Tool package.json Support: Custom Tools now support using package.json to define dependencies, automatically installing npm/bun packages, supporting any third-party libraries
• Support for customizing Plan Agent and Code Review Agent Models

Fixed

• Fixed bugs related to AI Pricing Service and LSP Tool
• Fixed API Usage Tab refresh issue
• Fixed OpenAI compatible type path handling (automatically appends /v1)

Acknowledgments

• Thanks to @qingzhongli for fixing OpenAI compatible type path handling
• Thanks to @MiaMia-Li for completing the "TalkCody Free Usage Guide" documentation

[0.3.0] - 2026-01-19

Added

• Google Login Support
• New create-agent command: One-click creation of custom AI agents
• New create-skill command: One-click creation of skills
• New create-tool command: One-click creation of custom tools
• New compact command: Manually compact conversations early
• Mac Keep Awake Support: Prevent system sleep
• Auto-Approve Plan Setting Support
• Reasoning-Effort Setting Support
• Commands Documentation

[0.2.6] - 2026-01-14

Added

• Agent Compatible with Claude Code Subagent Definition: Support for importing and using Claude Code subagent-defined agents from GitHub repositories, with more excellent built-in agents

Improved

• Task Title Generation Optimization: Improved task title generation logic for better user experience

Fixed

• Fixed Lint feature compatibility issue on Windows platform
• Fixed MiniMax Usage Cookie missing or expired issue
• Fixed Custom Tool refresh bug
• Fixed Chinese input method Enter key directly sending bug

Acknowledgments

• Thanks to @qingzhongli for fixing Lint feature compatibility issue on Windows platform

[0.2.5] - 2026-01-11

Added

• Custom Tools Support (Experimental): Added Custom Tools and Custom Tools Playground, allowing users to define custom AI tools with core capability to customize UI. For details, refer to Custom Tools
• Edit Auto-Approval: Can be enabled in settings, eliminating the need to wait for approval when modifying files
• Global file search now displays recently opened files
• When using multiple windows, the Project name is used as the window title

Improved

• UI Rendering Performance Optimization: Optimized UI rendering performance during multi-task parallel execution, improving response speed under complex workflows

Fixed

• Fixed File Changes Summary Bug: Fixed display issues with the file-changes-summary component
• Path Handling Compatibility: Fixed compatibility issues with Windows and Unix path handling
• Fixed Mac dock menu bug showing recent projects

[0.2.4] - 2026-01-06

Added

• Share Links: Support for generating exclusive links to share AI conversation tasks with others, with multiple expiration options and password protection. For details, refer to TalkCody Share Link Feature
• MiniMax Usage: Usage page now displays MiniMax Coding Plan usage
• Free Web Search: New free Web Search without additional configuration

Improved

• Plan Mode Optimization: Support for saving Plan results as files
• Global File Search Enhancement: Added support for hidden files and directories, with path search support
• Glob Tool Optimization: Support for searching hidden files and directories

Fixed

• Fixed External File Modification Detection: Resolved inaccurate detection when files are modified externally
• Fixed Keyboard Shortcut Conflict: Resolved Shift+Esc and Cmd+F keyboard shortcut conflict
• Fixed Agent Tool Timeout Issue: Optimized call-agent-tool timeout handling mechanism

[0.2.3] - 2026-01-03

Added

• Usage Page: New usage statistics page to view usage for Claude subscription, OpenAI subscription, GitHub Copilot subscription, and Zhipu Coding Plan.
• Free Usage Expansion: now supporting 8 completely free usage methods. For details, refer to TalkCody Free Usage Guide.
• GitHub Skill Import: Support for one-click import of Agent Skills from GitHub repositories.
• macOS Dock Menu Support: Support for quickly opening recent projects and creating new windows, enhancing the macOS user experience.
• Vue Syntax Highlighting Support: New Vue syntax highlighting and LSP support for a better Vue project development experience.

Improved

• Optimized Plan Agent: Improved Plan Agent execution logic.
• Agent Skills Specification Refactoring: Refactored Skills to Agent Skills Specification for better standardization of agents.
• Improved Timeout Handling: Optimized call-agent-tool timeout handling mechanism.

Fixed

• Fixed Linux Chinese Input: Resolved Chinese input method compatibility issues on Linux.

[0.2.2] - 2025-12-30

Added

• Unified Pages: Unified Explore page and Chat page for a more coherent user experience.

Improved

• Optimized Context Compaction: Improved message filtering and compaction logic to better support Long Run Tasks.
• Optimized GPT-5.2 Codex Experience: Optimized prompts for use via OpenAI subscriptions.
• Optimized Sub-Agent Calls: Improved timeout handling when calling sub-agents.

Fixed

• Fixed GitHub Copilot Image Handling: Resolved a bug in GitHub Copilot provider when handling image requests.
• Fixed Editor Synchronization: Fixed an issue where the editor could not perceive external file modifications in a timely manner.
• Fixed UI Display: Fixed a bug where the Diff display was incorrect for large files in the Edit File result UI.

[0.2.1] - 2025-12-27

Added

• Add Qwen Code Free Provider, to save your coding cost
• Add Github Copilot Free Provider, to save your coding cost

Improved

• Improve context compaction logic, to better handle long conversations

[0.2.0] - 2025-12-25

Added

• Claude model support via Claude Pro/Max subscription
• OpenAI model support via ChatGPT Plus/Pro subscription
• LSP Support: Complete Language Server Protocol implementation with code navigation, code completion, go-to-definition, find references, and other core features, significantly improving code understanding and editing experience
• Exa Free Web Search: New Exa search engine provider for out-of-the-box web search
• Custom Terminal Font: Support for custom terminal font and font size settings
• MiniMax 2.1 model support (free to use via TalkCody Provider)
• GLM 4.7 model support (TalkCody supports hot model updates, latest excellent coding models are usually auto-updated within 12 hours)

Improved

• Improved write file and edit file tool result display
• Optimized Prompt Cache for further cost savings

Fixed

• Fixed Windows terminal related bug
• Fixed model selector list duplicate display bug
• Fixed write file tool failing to write files correctly in some cases
• Fixed incorrect git status display in multi-window scenarios

[0.1.20] - 2025-12-19

Added

• MiniMax M2.1 beta model support - all users can try it for free. Select MiniMax M2.1 Beta model with talkcody provider to use it.

Improved

• Improved Plan Mode approval and Diff Review UI

[0.1.19] - 2025-12-18

Added

• Multi-Agent parallel execution (experimental) - multiple agents can now execute tasks simultaneously
• Git Worktree-based Task parallel execution (experimental) - tasks can run in parallel in isolated working directories
• One-click Git Commit: New Commit button in file changes summary with AI-generated commit messages
• One-click Code Review: New Review button in file changes summary to invoke Code Review Agent

Improved

• Improved MCP tool selection buttons
• Improved local agent loading performance
• Improved Edit File tool
• Improved Bash tool dangerous command detection
• Improved Context Compaction logic
• Improved AI request retry strategy

Fixed

• Fixed Windows terminal bug

[0.1.18] - 2025-12-13

Added

• Editor Lint feature with real-time syntax checking and quick fixes
• GPT 5.2 model support
• New feature documentation: Coding Plan, Web Search, Voice Input, Code Lint

Improved

• Improved Web Search tool
• Improved Web Fetch tool

Fixed

• Fixed Windows PowerShell terminal bug
• Fixed missing environment variables due to not correctly reading bashrc/zshrc

[0.1.17] - 2025-12-11

Added

• Support for multiple sessions running in parallel, significantly improving workflow efficiency
• Custom AI provider configuration support
• New built-in AI provider: Moonshot
• New built-in model: GLM-4.6 v
• New built-in Minimax Coding Plan MCP with web search and image input support
• New built-in GLM Coding Plan MCP with web search and image input support
• What's New dialog to show changelog after updates

Improved

• Optimized Bash tool output for better performance
• Optimized GitHub PR tool output for better performance
• Optimized Context Compaction logic for improved multi-turn conversation performance

Fixed

• Fixed HTTP MCP server header configuration bug
• Fixed Stdio MCP server not supporting custom environment variables
• Fixed database exit issue when using multiple windows