Skip to content
ReviewRamble

The weekly brief

AI, reviewed, in your inbox.

  • New reviews of tools, models, MCP servers and repos, with the verdict up front
  • Re-checked prices and ratings when something changes
  • New analysis pieces as they are published

Models, agents, MCP servers and the tools actually worth your time.

One email a week. Unsubscribe with the link in any brief. How we use your email

Taskmaster (claude-task-master)

CLI and MCP server that turns a PRD into a dependency-ordered task list your coding agent works through; quiet since April 2026.

Reviewed by ReviewRamble editors · Updated

Share this page:XLinkedInFacebookRedditWhatsAppEmail

Our rating

3.5/5

Taskmaster gives a coding agent a sensible plan to follow, but it loads 21,000 tokens of tools by default, sends prompt data to Sentry unless you opt out, and has gone quiet as its authors move to Hamster.

What is Taskmaster (claude-task-master)?

Taskmaster reads a product requirements document and breaks it into numbered tasks with dependencies, priorities, test strategies and subtasks, stored as JSON in a .taskmaster/ folder in your repository. Your coding agent then asks it for the next unblocked task, expands a big one into subtasks, marks work done and logs notes, through either the `task-master` CLI or an MCP server that Cursor, Windsurf, VS Code, Claude Code, Amazon Q and others can load.

The planning itself is done by a model you configure: a main model, an optional research model (Perplexity is the usual pick) and a fallback. It supports Anthropic, OpenAI, Google, Mistral, Groq, xAI, OpenRouter, Azure and Ollama keys, or the Claude Code and Codex CLIs with no separate key. Tags keep separate task lists for branches or workstreams, and a complexity report flags tasks worth splitting.

The authors, Eyal Toledano and Ralph Khreish, now build Hamster, a paid planning workspace, and Taskmaster's docs moved to tryhamster.com. A connected mode syncs tasks with Hamster for team features. The licence is MIT with the Commons Clause, which forbids selling Taskmaster itself or offering it as a hosted service.

Is Taskmaster (claude-task-master) worth it?

Taskmaster fixes a real failure of coding agents: on a project longer than one session they lose the plan. Writing a PRD and having Taskmaster turn it into ordered tasks with dependencies gives the agent a queue to work through, and `next` always returns something unblocked. That idea won it about 28,500 GitHub stars and 13,059 npm downloads in the week to 19 November 2026.

The cost lands in your context window. By default the MCP server loads all 36 tools, about 21,000 tokens, before the agent reads any of your code. The README's fix is TASK_MASTER_TOOLS: core mode loads 7 tools in about 5,000 tokens and standard mode 15 in about 10,000. Set it on day one.

Privacy needs a look too. In our reading of the 0.43.1 source, anonymous telemetry is on unless .taskmaster/config.json sets anonymousTelemetry to false, and it initialises Sentry with recordInputs and recordOutputs enabled for AI calls and sendDefaultPii set to true, so prompts built from your PRD and tasks can reach Sentry. The licence also deserves a correction: Hamster's product page calls Taskmaster MIT licensed, but the LICENSE file adds the Commons Clause, so you cannot sell it or run it as a hosted service.

The catch: development has slowed sharply. The last release, 0.43.1, came on 31 March 2026, and the last commits in April only moved documentation links to Hamster. Spec-kit, Beads and the planning modes now built into Claude Code, Cursor and Kiro cover much of the same ground, and the generated tasks are only as good as the PRD and the model you give it.

Our take: for a solo developer starting a new project from a written spec, Taskmaster in core mode with telemetry off is still a useful scaffold; for anything long-lived, weigh it against tools whose maintainers are still shipping.

Where it earns its place

  • A solo founder building an MVP with Cursor

    Stop the agent wandering off-plan between sessions

    Write the PRD to .taskmaster/docs/prd.txt, ask the agent to parse it, then work task by task with `next`. Run in core mode and set anonymousTelemetry to false; parsing a PRD costs a few model calls on your own key.

  • A Claude Code user on a Pro or Max subscription

    Plan tasks without buying separate API credits

    Set the main model to claude-code/sonnet so Taskmaster calls the local Claude Code CLI instead of an API key; add a Perplexity key only if you want the research command.

  • A two-person agency splitting client work across branches

    Keep separate task lists for a feature branch and the backlog

    Use tags to hold one task list per workstream and `move` with --with-dependencies to shift tasks between them. Teams that want shared, hosted planning are pointed to Hamster Team at $40 a creator a month.

Before you commit

Setup
`claude mcp add taskmaster-ai -- npx -y task-master-ai` or the Cursor one-click link, then add at least one provider key; or npm install -g task-master-ai and task-master init. Needs Node.js 20 or newer.
Integrations
MCP config for Cursor, Windsurf, VS Code, Amazon Q CLI, Claude Code and others; providers Anthropic, OpenAI, Google, Mistral, Groq, xAI, OpenRouter, Azure, Ollama, Perplexity, Claude Code CLI and Codex CLI; optional Hamster sync.
Data and privacy
Tasks live as JSON in .taskmaster/ in your repository and prompts go to the provider you choose. Sentry telemetry is on by default with AI inputs and outputs recorded; set anonymousTelemetry: false in .taskmaster/config.json. API keys sit in .env or your MCP config.
Cost at scale
No fee for the tool; each parse, expand or research call is a model request on your key. Hamster's hosted team features cost $40 a creator a month, or $400 a year.
Limits
36 tools and about 21,000 tokens by default; 160 open issues; no release since March 2026; the Commons Clause bars selling or hosting it as a service.

What’s good

  • Turns a PRD into tasks with dependencies, priorities and test strategies
  • Works as an MCP server in Cursor, Claude Code, VS Code and Windsurf, or as a CLI
  • Many model providers, including Claude Code and Codex CLIs without an API key
  • Tool-loading modes cut its context cost from about 21,000 to 5,000 tokens

What’s not

  • All 36 tools load by default, about 21,000 tokens
  • Telemetry to Sentry is on by default and records AI inputs and outputs
  • Commons Clause licence, not open source, despite MIT branding
  • No release since 31 March 2026; authors' focus has moved to Hamster
  • Task quality depends on a detailed PRD

Best for

Solo developers and small teams who write a PRD before building and want their coding agent to follow an explicit, dependency-ordered task list.

Taskmaster (claude-task-master) alternatives

Taskmaster (claude-task-master): common questions

Is Taskmaster (claude-task-master) worth it?
Taskmaster gives a coding agent a sensible plan to follow, but it loads 21,000 tokens of tools by default, sends prompt data to Sentry unless you opt out, and has gone quiet as its authors move to Hamster. We rate it 3.5 out of 5.
How much does Taskmaster (claude-task-master) cost?
Free to use under MIT with the Commons Clause (source-available, not open source); you pay the model provider. Optional Hamster workspace: Free with 10 briefs a month, Team $40/creator/month or $400/year, Enterprise $200/creator/month.
What are the downsides of Taskmaster (claude-task-master)?
All 36 tools load by default, about 21,000 tokens; telemetry to Sentry is on by default and records AI inputs and outputs; commons Clause licence, not open source, despite MIT branding; no release since 31 March 2026; authors' focus has moved to Hamster; task quality depends on a detailed PRD.
Who is Taskmaster (claude-task-master) best for?
Solo developers and small teams who write a PRD before building and want their coding agent to follow an explicit, dependency-ordered task list.
How much does Taskmaster (claude-task-master) cost for a team or in production?
No fee for the tool; each parse, expand or research call is a model request on your key. Hamster's hosted team features cost $40 a creator a month, or $400 a year.
Is Taskmaster (claude-task-master) safe for company data?
Tasks live as JSON in .taskmaster/ in your repository and prompts go to the provider you choose. Sentry telemetry is on by default with AI inputs and outputs recorded; set anonymousTelemetry: false in .taskmaster/config.json. API keys sit in .env or your MCP config.
What are the best Taskmaster (claude-task-master) alternatives?
The closest alternatives we have reviewed are Spec Kit, Beads, BMAD Method and Vibe Kanban.

Answers checked Nov 20, 2026.