Taskmaster
A task management system for AI-driven development that parses PRDs into tasks, usable as an MCP server in editors or as a CLI.
Tested on LinuxKnown vulnerabilities
Use it
Before you install: known vulnerabilities
pickfit found 15 known high or critical vulnerabilities (16 findings, one per affected package version) in the packages a fresh install brings (OSV.dev, 16 Nov 2026). Weigh this before you install it.
npm:@simple-git/argv-parser 1.1.1simple-git: `VISUAL` editor environment variable is omitted from unsafe editor detection (GHSA-v5rq-49vh-5v5c, critical, fixed in 2.0.1)npm:simple-git 3.36.0simple-git unsafe-operation guard does not block trailer command configuration (GHSA-x6jw-m9v5-85vh, critical, fixed in 4.0.1)
See 8 more
npm:@fastify/busboy 2.1.1— GHSA-x8mw-p69m-v3mx (high), fixed in 3.2.1npm:@opentelemetry/exporter-prometheus 0.203.0— GHSA-q7rr-3cgh-j5r3 (high), fixed in 0.217.0npm:@opentelemetry/propagator-jaeger 2.0.1— GHSA-45rx-2jwx-cxfr (high), fixed in 2.9.0npm:@opentelemetry/sdk-node 0.203.0— GHSA-q7rr-3cgh-j5r3 (high), fixed in 0.217.0npm:extract-zip 2.0.1— GHSA-7pqw-9j4j-h8q3 (high), no fix yetnpm:extract-zip 2.0.1— GHSA-jmr9-qjv8-65gv (high), no fix yetnpm:simple-git 3.36.0— GHSA-858h-whjf-mvg5 (high), fixed in 4.0.0npm:simple-git 3.36.0— GHSA-g4wm-2vf7-vfgr (high), fixed in 4.0.0
6 more findings not listed here.
Linux x86_64
You need first
- Needs the network
- Keys or accounts: At least one AI provider API key (e.g. Anthropic, OpenAI, Google Gemini, Perplexity, xAI, OpenRouter)
These steps ran from a folder named work in the home folder, which already existed. Make it and run them from there:
mkdir -p "$HOME/work" && cd "$HOME/work"npm install -g task-master-ai
npm install -g task-master-ai
Verified in pickfit: the commands exactly as its sandbox ran them
For the record only: they name the sandbox's home, /home/sbx.
npm install -g task-master-aitask-master --versionmkdir -p /home/sbx/work/proj/.taskmaster/tasks && cat > /home/sbx/work/proj/.taskmaster/tasks/tasks.json <<'EOF'
{
"master": {
"tasks": [
{"id": 1, "title": "Set up project skeleton", "description": "Create repo layout", "details": "", "testStrategy": "", "status": "done", "dependencies": [], "priority": "high", "subtasks": []},
{"id": 2, "title": "Implement login API", "description": "JWT based login endpoint", "details": "", "testStrategy": "", "status": "pending", "dependencies": [1], "priority": "high", "subtasks": []},
{"id": 3, "title": "Write docs", "description": "Document the API", "details": "", "testStrategy": "", "status": "pending", "dependencies": [2], "priority": "low", "subtasks": []}
],
"metadata": {"created": "2024-01-01T00:00:00.000Z", "updated": "2024-01-01T00:00:00.000Z", "description": "Tasks for master context"}
}
}
EOFcd /home/sbx/work/proj && task-master listCheck that it works
task-master --versionThe output should contain “0.43.1”.
First thing to try
mkdir -p "$HOME"/work/proj/.taskmaster/tasks && cat > "$HOME"/work/proj/.taskmaster/tasks/tasks.json <<'EOF'
{
"master": {
"tasks": [
{"id": 1, "title": "Set up project skeleton", "description": "Create repo layout", "details": "", "testStrategy": "", "status": "done", "dependencies": [], "priority": "high", "subtasks": []},
{"id": 2, "title": "Implement login API", "description": "JWT based login endpoint", "details": "", "testStrategy": "", "status": "pending", "dependencies": [1], "priority": "high", "subtasks": []},
{"id": 3, "title": "Write docs", "description": "Document the API", "details": "", "testStrategy": "", "status": "pending", "dependencies": [2], "priority": "low", "subtasks": []}
],
"metadata": {"created": "2024-01-01T00:00:00.000Z", "updated": "2024-01-01T00:00:00.000Z", "description": "Tasks for master context"}
}
}
EOFcd "$HOME"/work/proj && task-master listAdd it to your agent
claude mcp add claude-task-master -- task-master-aiJSON for Claude Desktop, Cursor and other clients
{
"mcpServers": {
"claude-task-master": {
"command": "task-master-ai",
"args": []
}
}
}Inside a project folder that has a .taskmaster directory, run task-master list to see your tasks, or add task-master-ai as an MCP server in your editor.
What it does
- Get a suggestion for the next task to work onTested
How to use it
- Keep your tasks in .taskmaster/tasks/tasks.json inside your project folder.
- Connect your editor's MCP client to the
task-master-aiserver. - Call the next_task tool with your project folder as projectRoot, for example
{"projectRoot": "<project folder>"}.
- Turn a PRD into a list of tasksNot tested
- Break a task into subtasksNot tested
- Research current information using project contextNot tested
How far pickfit got
Linux x86_64
- Installs
- Connects
- 1 of 4 features work
- Works offline
macOS and Windows: the project says it runs there; pickfit hasn’t tried it.
Good for
- generate tasks from a PRD
- find the next task to work on
- research current best practices with project context
How pickfit tested it
Each run starts from a clean machine and follows the steps above. How pickfit verifies
Linux x86_64
An MCP client called one of its tools
- Replayed from scratch
- Works offline
- Checked on 15 Nov 2026
- Version 0.43.1
An MCP client handshake over stdio listed 44 tools, and a call to next_task returned a result.
Differences the test machine hit
- README: Launch with npx -y task-master-ai
In the test: After npm install -g task-master-ai, the task-master-ai binary starts the MCP server directly over stdio (no npx needed) - README: MCP server speaks JSON-RPC over stdio
In the test: The server printed 1 non-JSON-RPC line on stdout; strict MCP clients may complain - README: Parse PRD / expand / research via AI chat
In the test: These tools call an AI provider and need an API key; only non-AI tools (next_task, list) work without one - README: An MCP server over stdio writes only JSON-RPC messages to stdout
In the test: This server wrote 1 other line(s) to stdout during the handshake; strict MCP clients may reject them.