NeuroDock Task Fractionator — Claude Code MCP server
Turns a vague goal into 5–90 minute tasks and selects one safe next action. Use it when an MCP client needs a small, ordered work plan.
claude mcp add neurodock-mcp-task-fractionator -- uvx neurodock-mcp-task-fractionatorWhat NeuroDock Task Fractionator does
NeuroDock Task Fractionator exposes two MCP tools. decompose turns a plain-language goal into atomic tasks with acceptance criteria, dependency order and tags. next_one returns exactly one pending task, plus a short rationale and a confidence score.
The server runs locally, makes no remote calls and contains no LLM call. Goal text is never written to logs. Decomposition results are stateless, while the next-task tool reads pending work through a PendingTaskSource.
Version v0.0.3 defaults to the in-memory source, so tasks must be seeded in the running process before next_one can select one. The graph source is still a Phase 2 stub and returns COGNITIVE_GRAPH_UNAVAILABLE, which is a concrete limit for persistent task storage.
What it helps with
- break a vague goal into 5–90 minute tasks
- attach acceptance criteria to each task
- order work around dependencies
- select one unblocked next action
Example prompts for the agent
- Break “prepare the service release” into atomic tasks and fit them into
PT4H. - For project
founding-scope-rfc, choose the single next task. - Turn this initiative into dependency-ordered tasks with acceptance criteria.
How to install NeuroDock Task Fractionator
Platform CLI
claude mcp add neurodock-mcp-task-fractionator -- uvx neurodock-mcp-task-fractionatoruvx runs the command named after the package; if the README names another one, use uvx --from neurodock-mcp-task-fractionator <command>
Platform CLI
codex mcp add mcp-task-fractionator -- uvx neurodock-mcp-task-fractionatoruvx runs the command named after the package; if the README names another one, use uvx --from neurodock-mcp-task-fractionator <command>
Config file
{
"mcpServers": {
"mcp-task-fractionator": {
"command": "uvx",
"args": [
"neurodock-mcp-task-fractionator"
]
}
}
}uvx runs the command named after the package; if the README names another one, use uvx --from neurodock-mcp-task-fractionator <command>
Config file
{
"servers": {
"mcp-task-fractionator": {
"type": "stdio",
"command": "uvx",
"args": [
"neurodock-mcp-task-fractionator"
]
}
}
}uvx runs the command named after the package; if the README names another one, use uvx --from neurodock-mcp-task-fractionator <command>
Platform CLI
gemini mcp add mcp-task-fractionator uvx neurodock-mcp-task-fractionatoruvx runs the command named after the package; if the README names another one, use uvx --from neurodock-mcp-task-fractionator <command>
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about NeuroDock Task Fractionator
How do I install NeuroDock Task Fractionator in Claude Code?
Run claude mcp add neurodock-mcp-task-fractionator -- uvx neurodock-mcp-task-fractionator. Claude Code will connect to the server over MCP stdio.
What does NeuroDock Task Fractionator do?
It decomposes a goal into atomic 5–90 minute tasks, then can select one pending task whose dependencies are complete.
Why can’t next_one use persistent tasks?
In v0.0.3, memory is the default source and keeps tasks only in process memory. The graph source is a Phase 2 stub that returns COGNITIVE_GRAPH_UNAVAILABLE.
