# tlennon-ie/neurodock [Health: Active]

**Category:** 💬 Communication  
**Repository:** https://github.com/tlennon-ie/neurodock  
**GitHub Stars:** 17  
**npm Downloads (last month):** 1465  
**Views:** 2  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/tlennon-ie-neurodock

## Description
Local-first cognitive substrate for neurodivergent professionals: a translator for corporate ambiguity, a planner, and a guardrail that declines to amplify rumination, hyperfocus, or sycophancy. Hosted stateless tools over OAuth.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "neurodock": {
    "command": "npx",
    "args": ["-y","@neurodock/cli@latest"]
  }
}
```

## Documentation

## What tlennon-ie/neurodock MCP server does

tlennon-ie/neurodock MCP server provides a set of MCP servers aimed at neurodivergent professionals. Its capabilities cover remembering work context, interpreting time and pacing, breaking goals into smaller tasks, translating ambiguous corporate language, and declining to reinforce rumination or sycophancy. Self-identification is sufficient; the project does not require a diagnosis.

The local installation is composed of five servers: chronometric, cognitive-graph, task-fractionator, translation, and guardrail. An additional evals harness is installed alongside them. Claude remains the conversational interface while the MCP servers read and write the local NeuroDock store.

## How it works

By default, the components run on the user’s computer and communicate with an MCP-aware client over stdio. Local data is stored under `~/.neurodock`, and a starter profile is placed at `~/.neurodock/profile.yaml`.

The project also offers stateless remote versions of the translation, guardrail, and decomposition capabilities. These are available at `https://mcp.neurodock.org/mcp` and use OAuth rather than a local installation. The remote option does not provide the complete local five-server arrangement described above.

## Setup and configuration

The recommended installation uses Node’s package runner:

```sh
npx --yes @neurodock/cli@latest install-all
```

This installs the Python packages, configures Claude Desktop, Claude Code, or Cursor, and creates the starter profile. Python 3.11 or newer and Node 22 or newer are required. macOS, Linux, and Windows are supported. After installation, the client must be fully quit and restarted so it reloads its MCP configuration; closing only the window is not sufficient.

A manual route is also documented: install the five MCP packages and the evals package with `pip`, then run the CLI’s `init` command to write client entries. The CLI is distributed as `@neurodock/cli` on npm, while the MCP servers are distributed through PyPI.

Optional proactive guardrails can be installed with `neurodock install-hooks --self-test`. Claude Code hooks inspect tool-call patterns and emit a dismissible stderr notice without blocking the action. Adding `--install-daemon` installs a per-user background entry that polls every five minutes and can show native operating-system notifications. Guardrails can be disabled with `NEURODOCK_GUARDRAILS=off`.

## Tools and capabilities

The tlennon-ie/neurodock MCP server collection includes:

- Chronometric support for time awareness and pacing.
- Cognitive-graph storage for remembering work context.
- Task fractionation for turning goals into atomic tasks.
- Translation for corporate ambiguity.
- Guardrails that avoid amplifying rumination, hyperfocus, or sycophancy.

The repository also documents a Claude Code plugin, Claude Desktop extension bundles, and MCP Registry entries for the same local stdio servers. The hosted endpoint is an alternative for the stateless subset rather than a replacement for local memory.

## Limitations and notes

The local configuration depends on a client that supports MCP. The README specifically documents Claude Desktop, Claude Code, and Cursor. Claude must be fully restarted after configuration changes. Local-first behavior is the default, but explicitly enabled cloud options can change where data is processed. Proactive hooks are optional, and the daemon requires the separate hook installation step.

_Full upstream README: https://allmcps.com/mcp/tlennon-ie-neurodock/readme_