
Task Master AI
AI-powered task-management system for AI-driven development that drops into Cursor, Windsurf, Claude Code, and more.
Add to your client
Copy the config for your MCP client and paste it into its config file.
npx -y task-master-aiPaste into ~/Library/Application Support/Claude/claude_desktop_config.json
{
"mcpServers": {
"task-master-ai": {
"command": "npx",
"args": [
"-y",
"task-master-ai"
],
"env": {
"ANTHROPIC_API_KEY": "YOUR_ANTHROPIC_API_KEY_HERE",
"PERPLEXITY_API_KEY": "YOUR_PERPLEXITY_API_KEY_HERE",
"OPENAI_API_KEY": "YOUR_OPENAI_KEY_HERE",
"GOOGLE_API_KEY": "YOUR_GOOGLE_KEY_HERE",
"MISTRAL_API_KEY": "YOUR_MISTRAL_KEY_HERE",
"GROQ_API_KEY": "YOUR_GROQ_KEY_HERE",
"OPENROUTER_API_KEY": "YOUR_OPENROUTER_KEY_HERE",
"XAI_API_KEY": "YOUR_XAI_KEY_HERE",
"AZURE_OPENAI_API_KEY": "YOUR_AZURE_KEY_HERE",
"OLLAMA_API_KEY": "YOUR_OLLAMA_API_KEY_HERE"
}
}
}
}Step-by-step guides: Add to Claude Desktop · Add to Cursor · Add to Windsurf
Before you start
- Node.js with npx available
- An MCP-compatible editor/client (Cursor, Windsurf, VS Code, Claude Code, or Q CLI)
- At least one AI provider API key (Anthropic, OpenAI, Google, Perplexity, xAI, OpenRouter, etc.) OR Claude Code CLI / Codex CLI for keyless use
About Task Master AI
Taskmaster is an MCP server (and CLI) that turns a Product Requirements Document into an actionable, dependency-aware task list and helps an AI assistant work through it task by task. It exposes 36 MCP tools by default (configurable down to a standard/core subset to save context), supports a wide range of AI providers, and integrates with editors via the standard mcpServers/servers config. Initialize a project, optionally add a PRD at .taskmaster/docs/prd.txt, then ask your assistant to parse it, plan the next task, expand tasks, and research fresh information with project context.
Tools & capabilities (15)
initialize_projectInitialize Task Master in a project, creating the necessary files and structure.
parse_prdParse a Product Requirements Document (PRD) and generate structured tasks from it.
get_tasksList all tasks in the project.
get_taskShow details for a specific task (supports retrieving individual tasks by ID).
next_taskDetermine the next task to work on based on dependencies and status.
set_task_statusSet or update the status of a task.
add_taskAdd a new task to the project.
remove_taskRemove a task from the project.
add_subtaskAdd a subtask under an existing task.
update_subtaskUpdate an existing subtask with new information or progress.
expand_taskExpand a task into more granular subtasks.
expand_allExpand all eligible tasks into subtasks in one operation.
analyze_project_complexityAnalyze project/task complexity to inform planning and expansion.
complexity_reportProduce a complexity report for the project's tasks.
generateGenerate task files from the task data.
What this server can do
Task Master AI provides tools for these capabilities — tap one to see every MCP server that does the same:
When to use it
- Parse a PRD into a structured, dependency-aware task list for AI-driven development
- Ask your AI assistant what the next task to work on is and have it implement it
- Expand high-level tasks into detailed subtasks and track their status
- Analyze task complexity to plan work and prioritize
- Run research queries that incorporate your project's code context (e.g. migration strategies, best practices)
- Manage tasks across tags/workstreams (e.g. move tasks from backlog to in-progress)
Security notes
Requires at least one AI provider API key (e.g. ANTHROPIC_API_KEY, OPENAI_API_KEY, PERPLEXITY_API_KEY) supplied via the MCP config's env section or a project .env file. Claude Code and Codex CLI can be used without an API key. Keep your API keys secret; remove any provider keys you don't use from the config.
Task Master AI FAQ
Do I need an API key?
You need at least one AI provider API key (e.g. Anthropic, OpenAI, Perplexity) set in your MCP config's env section or a project .env file. Alternatively, you can use Claude Code or Codex CLI, which require no API key (Claude Code uses your local Claude instance; Codex uses OAuth via a ChatGPT subscription).
How do I reduce the context window used by the MCP tools?
Set the TASK_MASTER_TOOLS environment variable in your MCP config. Options are 'all' (36 tools, default), 'standard' (15 tools), 'core'/'lean' (7 tools), or a comma-separated list of specific tools. Core mode reduces token usage by roughly 70%.
Which editors are supported?
Cursor, Windsurf, VS Code, Claude Code, and Q CLI are documented. Cursor/Windsurf/Q use the 'mcpServers' key; VS Code uses 'servers' with a 'type': 'stdio'. There's also a one-click install link for Cursor 1.0+ and a 'claude mcp add' command for Claude Code.
What's the recommended starting point?
Start with a detailed PRD at .taskmaster/docs/prd.txt (an example template is generated on init), then ask your assistant to parse it. For complex projects a PRD is recommended, but you can also create individual tasks directly via chat.
Alternatives to Task Master AI
Compare all alternatives →Background computer-use MCP server that drives native macOS, Windows, and Linux desktop apps without stealing focus.
Self-hosted MCP server for Jira and Confluence Cloud and Server/Data Center.
Create, read, and modify Excel workbooks with your AI agent — no Microsoft Excel required.
Compare Task Master AI with: