RehabTools

Privacy & data handling

This page covers the RehabTools website and Rehabtools evidence plugin, operated by Ross Clark. Plugin information added 12 October 2026.

Rehabtools plugin in ChatGPT

The plugin offers read-only publication, search and fetch tools for already-public rehabilitation evidence. It does not provide patient-record storage, private document uploads, purchases or access to private drafts. The connected MCP endpoint is a fixed Netlify preview snapshot rather than an automatic feed of later website changes.

When ChatGPT invokes a tool, the tool request is sent to the RehabTools MCP endpoint hosted on Netlify. Requests may contain a search query, public record or section identifiers, a publication release identifier and pagination cursors. Search queries can include words from your prompt. The endpoint returns public evidence excerpts, citations and provenance to ChatGPT.

Content in your ChatGPT conversation is also handled by OpenAI under its policies and your account settings. Do not assume that a read-only tool means your question stays on your device. Hosting and network providers process the operational information needed to serve requests, such as connection and request metadata, and may retain logs under their policies. This page does not promise zero retention, a fixed log-deletion period or a particular processing country.

The tool interface does not provide a function for storing or deleting personal records. Clearing a conversation or browser page does not by itself delete information retained by OpenAI, Netlify or other providers. For their privacy controls and access or deletion requests, use the relevant provider's account settings and privacy-request process.

Use the chatbot for general educational guidance only. It is not a confidential channel to the operator, and it does not submit a privacy request to the operator. Do not submit patient, participant, student, identifying, confidential or other sensitive information.

OpenAI privacy policy · Netlify privacy policy

Website services

The existing website-specific information below is retained. The on-device site guide, hosted site guide and ChatGPT plugin use different data paths.

AI site guide

The on-device option downloads LFM2.5-350M model files from Hugging Face and runtime files from the jsDelivr CDN, and runs them locally. Questions and retrieved source excerpts remain on your device. Model files may remain in the browser cache; unloading releases the running model, while clearing site data removes cached files. This release requires WebGPU; CPU mode is not supported. GPU support and available memory vary by device. Switching to hosted AI is an explicit choice.

Using the hosted AI guide sends your question, up to three earlier questions and selected public site excerpts through Netlify to OpenRouter and a model provider. It does not receive uploaded files, SeeSway traces or your other browsing activity. Use general questions only; never enter personal or participant information.

The site does not save chat content to a database or browser storage. It remains in page memory until you clear it or reload. The hosted guide uses free models through NVIDIA, Google AI Studio or Liquid. These American providers may process requests outside the US. Zero data retention is not required: providers may retain questions or use them for training under their own policies. No paid model fallback is allowed; if free models are unavailable, the guide shows source links. Hosting and API services still process operational metadata under their own policies. These controls are not a guarantee that visitors cannot enter sensitive information.

OpenRouter privacy policy ↗

Your research data

SeeSway processes CSV data locally in this browser. Its analysis code does not upload trace values or filenames to a server. The runtime and scientific libraries download from this site on first use.

External applications

The three research companions now run from RehabTools. The allocation app uses recorded fictional cases and does not make live AI calls. Historical embedded teaching materials still load from their named external hosts; those services’ privacy practices apply.

Analytics and hosting

This version does not add visitor analytics, advertising trackers or tracking cookies. The hosting provider receives the network information needed to serve pages and may retain operational logs under its own policy.

Usage analytics is a planned improvement, not an active feature. Any implementation should exclude search text, filenames, uploaded data and participant information.

Statement updated 20 September 2026.

For general help, see limited support. See also the terms of use.