Task Researcher

What is Task Researcher?

Task Researcher is a Python task management system designed for AI-driven development. It integrates in-depth research capabilities using the knowledge-storm library to break down complex projects, generate tasks, and leverage automated research to inform implementation details. It provides both a command-line interface (CLI) and a Model Context Protocol (MCP) server.

How to use Task Researcher?

Install via clone and Poetry, configure a .env file with LLM and search API keys, then use the task-researcher CLI command or run task-researcher serve-mcp to start the MCP server over stdio transport. Clients like Claude Desktop connect by editing their claude_desktop_config.json.

Key features of Task Researcher

• Parse specifications to generate initial tasks.
• AI-powered task expansion into subtasks.
• STORM-based research workflow for complex tasks.
• Task update and dependency management.
• Complexity analysis with research hints.
• Standalone research report generation on any topic.

Use cases of Task Researcher

• Decomposing software project specifications into executable task lists.
• Generating detailed subtask breakdowns informed by web research.
• Updating existing task plans when requirements change.
• Analyzing and reporting task complexity for project planning.
• Producing in-depth research reports on technical topics using STORM.

FAQ from Task Researcher

What is the difference between --research and --research-hint?

expand --research triggers the full STORM-based workflow, performing live web searches to inform subtasks. analyze-complexity --research-hint modifies the prompt to ask the primary LLM to leverage its internal knowledge more deeply without any live search.

What are the runtime requirements?

Python 3.10+, an API key for at least one supported LLM provider (e.g., Anthropic, Google Gemini, OpenAI), and the knowledge-storm library. A search engine API key is required for the --research feature and the research-topic command.

Where does the task data live?

Task data is stored locally in a tasks/tasks.json file following Pydantic models. The file includes a meta section and a list of task objects with dependencies, phases, and subtasks.

How are API keys configured?

All credentials are set in a .env file. Required variables include LLM_MODEL, the provider's API key (e.g., ANTHROPIC_API_KEY), STORM_RETRIEVER, and the corresponding search API key (e.g., BING_SEARCH_API_KEY).

What transport does the MCP server use?

The MCP server uses stdio transport by default. It exposes tools like parse_inputs, expand_task, and research_topic, as well as resources such as tasks://current and report://complexity.