Changes

Author reusable skills that shape how Hamster approaches your team's work.

Connect GitHub, Linear, and more during onboarding. Synced tasks now show assignees who aren't in your workspace yet.

Shared brief links now show a preview page. Hamster now also revises plans mid-execution.

The redesigned Activity tab groups votes, status changes, and comments by document version.

The redesigned Plan tab shows each phase of generation, plus Figma integration for briefs and blueprints.

Your AI coding agent can now set up an entire project in Hamster from a single conversation.

Inline diffs now highlight only the changed words, not entire paragraphs.

The redesigned Context tab consolidates files, URLs, and integrations into one unified view.

Two-way sync with Linear, plus guided onboarding that adapts to your workflow.

The CLI and MCP server give your coding agents live access to briefs, tasks, and plans. Plus document version history and six new languages.

A new feedback widget lets you submit reports without leaving your workflow.

Navigation shortcuts, recently viewed items, and keyboard shortcuts in the command palette.

Structured questions now appear throughout the entire product, with automatic clarification for ambiguous answers.

Chat directly with Hamster to add, remove, or restructure tasks in your plan.

Plans now generate in a dedicated thread where you can watch progress and drill into any task.

Hamster now researches topics independently and surfaces findings in your thread.

Structured discovery questions let Hamster collect the right information step by step.

Paste a Google Drive, Notion, Jira, or GitHub URL in chat and Hamster queries your connected integration directly.

Blueprints and Methods now have a full mobile chat interface with tab navigation.

Link your Google Drive account and Hamster pulls in your documents when generating briefs and blueprints.

Ask Hamster to look into a competitor or industry trend and it returns structured findings for your briefs.

Tasks now support a custom metadata field for storing external IDs and workflow data, and Taskmaster can be installed in Claude Desktop in one click.

Chat history is now used as context when Hamster processes your requests.

The new block-based editor supports real-time collaboration, change previews, and inline images.

Hamster now pauses mid-workflow to ask for clarification or get your approval before proceeding.

New --ready and --blocking filters help identify parallelizable and high-impact tasks. Plus cross-process file locking prevents data loss with multiple Claude Code windows.

A step-by-step wizard walks you through workspace setup, team invites, and uploading company context.

New tm loop command runs Claude Code in automated iterations, executing tasks from your backlog with built-in presets for TDD, linting, duplication cleanup, and more.

Connect your Slack workspace so Hamster can reference team discussions when generating briefs.

Blueprints and Methods give Hamster structured knowledge about your company and team processes.

CLI startup speed improved by 2x, smarter project root detection with boundary markers for monorepos, and reasoning effort validation for OpenAI models.

Real-time task monitoring with watch mode, auto-detection of 13 IDEs for rules setup, compact list output, and Gemini 3 Flash Preview model support.

Tasks in team mode now display rich implementation guidance including relevant files, coding patterns, and acceptance criteria. Plus deferred MCP loading saves 16% of Claude Code's context window.

The diff viewer now accurately shows proposed changes for documents of any structure or size.

Task Master slash commands now available for Roo, Cursor, Codex, Gemini, and OpenCode. Solo and team modes get distinct command sets with automatic cleanup when switching modes.

Fixed login issues for users with browser/firewall restrictions and added tm/ prefix to autopilot branch names for better team organization.

Improved authentication error handling and increased page size for brief selection in the interactive CLI setup.

Deep Hamster integration with parse-prd, new tm export command for team collaboration, redesigned tm init workflow, and simpler positional command syntax.

Chat threads auto-generate titles based on conversation content.

The generate command is back by popular demand with Markdown output format and tag filtering. Plus Hamster integration rules for the rules system.

Upgraded Gemini CLI provider to native structured output support with simplified implementation, and added support for Opus 4.1 and Opus 4.5 Anthropic models.

The generate command has been marked as deprecated in preparation for the new Markdown-based file generation system.

Added Gemini 3 Pro Preview to supported Task Master AI providers for both Google and Gemini CLI.

Upgraded fastmcp dependency to resolve MCP server compatibility issues where clients reported 'Server does not support completions'.

Fixed visual alignment in CLI box titles by adding emoji variant selector to warning sign.

All AI providers now support HTTP/HTTPS proxies for corporate environments, restored JSON output for list/show commands, and Claude Code provider now respects your configuration files.

New users are guided through setup before accessing the full platform.

Improved schema compatibility across LLM models for parse-prd, fixed pricing display after JSON repair, and enabled structured outputs for Z.ai providers.

Added separate Z.ai provider for Coding Plan subscription endpoint and improved auto-update experience with automatic restarts.

Connect Task Master to any OpenAI-compatible service, use high-performance Z.ai GLM models with 200K+ context windows, or run completely offline with LM Studio at zero API cost.

Realtime Replay lets you step through every change to a document and see who made it.

Introducing autonomous test-driven development with `tm autopilot` commands that guide AI agents through complete RED-GREEN-COMMIT cycles. Plus configurable MCP tool loading to reduce AI context usage by 80% with lean mode.

Connect your Linear workspace so Hamster can reference your issues and projects.

Launch of the official Claude Code plugin with 49 slash commands and 3 AI agents, plus the new RPG (Repository Planning Graph) method for creating dependency-aware PRDs that generate perfectly ordered task graphs.

Unlock GPT-5 and GPT-5-Codex models with 272K context through OAuth-based Codex CLI integration. Major AI SDK v5 migration brings structured output generation and improved reliability across all AI providers.

Fixed critical issue where `tm show` command couldn't find subtasks using dotted notation IDs (e.g., '8.1'), restoring proper subtask navigation and information display.

Improved subtask status handling: parent tasks automatically update to 'done' when all subtasks complete, and to 'in-progress' when any subtask is active or completed.

Major automation release: `tm start` auto-launches Claude Code with comprehensive task context, seamless background auto-updates, Grok CLI support with full codebase context, smart parent task status management, and cross-tag task movement for organized development workflows.

Enhanced Gemini CLI with automatic codebase analysis for context-aware task generation, plus restored Claude Code commands with improved command organization and configurable codebase analysis features.

Prevented crashes when Claude Code SDK is missing through improved error handling and SDK presence checks. Updated dependency to v1.0.88 for better export consistency.

Organize tasks across different contexts with cross-tag movement functionality, plus new Kilo Code profile integration with custom modes and a compact list view for streamlined task management.

Automatic codebase analysis for Claude Code provider in parse-prd, expand-task, and analyze-complexity commands. AI now analyzes project structure, existing implementations, and patterns before generating tasks or subtasks.

Official TaskMaster extension now available on VS Code Marketplace and Open VSX Registry, bringing task management directly into your development environment.

Full Kiro IDE support with 7 autonomous task management agent hooks, plus intelligent prompting to generate complexity reports when missing for better task planning.

Added support for Kiro IDE with custom rule files, MCP configuration, and automatic generation of rule files in .kiro/steering/ directory with markdown format.

Added support for xAI's Grok 4 model with competitive $3/$15 per 1M token pricing, available for main, fallback, and research roles.

Support for gemini-cli provider with free/subscription Google Accounts and Gemini Cloud Assist. AI can now determine optimal task counts based on complexity analysis, plus advanced Claude Code settings.

Configure custom base URLs for AI providers using environment variables like OPENAI_BASE_URL, enabling use of proxy services and alternative endpoints.

Fixed critical contextGatherer bug that caused crashes when adding tasks, resolving 'Cannot read properties of undefined' error for improved stability.

Comprehensive research command with intelligent task discovery through fuzzy search, multi-source context gathering from tasks/files/project structure, and interactive follow-ups for deep project insights.

Improved findTasks algorithm for better path resolution, fixed MCP update tool 'No valid tasks found' error, and enhanced add-task fuzzy search for improved user experience.

Resolved AWS Bedrock integration issues, fixed MCP tool call logging errors, and updated rules for improved directory structure compatibility.

Added support for AWS Bedrock and Google Vertex AI providers, expanding AI model options with enterprise-grade cloud AI services for enhanced task management capabilities.

Enhanced Ollama integration with live model validation against local instances, interactive setup with custom model options, and real-time validation through /api/tags endpoint queries.

Added Ollama as a supported AI provider with interactive setup, plus per-role base URL configuration (main, research, fallback) allowing custom endpoint overrides for any provider.

Fixed critical ERR_MODULE_NOT_FOUND error when starting MCP Server, ensuring reliable server initialization and module resolution.

Intelligent task expansion that automatically uses complexity analysis to determine subtask counts and tailored prompts. Default append behavior with --force flag for fresh starts, eliminating manual prompt copying.

Streamlined MCP server access through npx task-master-ai command, replacing complex setup with simple one-command execution for easier integration with development tools.

Choose your workflow: install as a command-line tool (task-master) or use as MCP server within IDEs like Cursor. Includes new initialize_project capability for simplified project setup in integrated tools.

Complete MCP server implementation for all commands using tools, plus automated GitHub Actions for streamlined releases and package management improvements.

Early foundation release of TaskMaster with core task management functionality. See full details at npmjs.com/package/task-master-ai/v/0.9.30.