Taskmaster
A task manager for AI-driven development: breaks a PRD into dependent tasks and guides the agent through them via MCP or CLI
Install
claude mcp add taskmaster-ai -- npx -y task-master-aiPut provider keys in the project .env or the env section of the MCP config.
This is third-party code. Review the repository files before installing.
What it does
Taskmaster turns a requirements document into a structured task list with dependencies and subtasks. The agent picks the next available task, completes it and updates the status, so a long project survives across sessions. It runs as an MCP server in Cursor, Windsurf, VS Code, Claude Code and other editors, or as the task-master CLI. You can set main, research and fallback models from different providers, split work by tags and use a research mode.
Who it is for. For developers running projects with an agent that span more than one session.
Good fit when
- You have a PRD and need agent-ready tasks
- The agent loses track in a long project
- You need to see task dependencies and order
Not a fit when
- The task fits in a single request
- Tasks already live in a tracker the agent works with
Example request
Parse my PRD into tasks and tell me which one to do nextLimitations
Generating tasks needs a key for at least one provider: Anthropic, OpenAI, Gemini, Perplexity, xAI or OpenRouter; Claude Code or Codex CLI work without a key. Detailed docs live on tryhamster.com, and the license is non-standard.
How to disable. Remove the task-master-ai server from your MCP config, the package with npm uninstall -g task-master-ai and, if needed, the project's .taskmaster folder.
MCP
- Transport
- stdio
- Authentication
- API key
| Environment variables | |
|---|---|
| ANTHROPIC_API_KEY secret | Anthropic key; at least one provider key is required |
| OPENAI_API_KEY secret | OpenAI key |
| PERPLEXITY_API_KEY secret | Perplexity key for the research model |
| OPENROUTER_API_KEY secret | OpenRouter key |
| TASK_MASTER_TOOLS | Tool set: all, standard, core or a comma-separated list |
Security check
- Sends project descriptions to external LLM providers
- Stores API keys in config
README in short
The README links to the docs and gives a one-click Cursor install and a Claude Code command. It lists key requirements and the three model roles. The quick start shows MCP config paths for editors and JSON snippets with env vars, then the CLI path: init, parse-prd, list and next.
FAQ
MCP or CLI?
The README recommends MCP for editor use. The CLI covers task-master init, parse-prd, list and next in a terminal.
How many keys do I need?
At least one. The research model is optional but recommended.
Related
A CLI for every Google Workspace API with JSON output and agent skills: Drive, Gmail, Calendar, Sheets and more
Local search over Markdown notes, docs and meeting transcripts: keywords, semantic search and reranking, with an MCP server
Anthropic's official plugins for knowledge work roles: sales, support, product, marketing, legal, finance, data and enterprise search
An MCP server for Jira and Confluence, cloud and self-hosted: search, create and edit issues and pages