NeuroDock — Translate

Decode corporate subtext and check tone before you send. Local-first by default; cloud only with your explicit consent.

NeuroDock is a translator for work communication, built for neurodivergent professionals, ADHD, autistic, AuDHD, OCD, dyslexic, and every combination. It works on the sites you already use, on the text you select, on demand. It does not change your messages without asking. ## Why it helps - "I stopped re-reading the same Slack message ten times trying to figure out if my manager was annoyed." - "I check the tone of every important reply before I hit send. It saved me from at least three apologies last month." - "It runs on my laptop. None of my work email goes through anyone else's server." ## What it does Four actions, on demand, on text you select or compose: 1. **Translate incoming** — surfaces the literal subtext under a vague Slack message or a polite-sounding email, anchored to specific spans of the original so you can see exactly what triggered each read. 2. **Check tone** — reads back a draft you are about to send and tells you how it will land, with concrete suggestions when the read is off. 3. **Rewrite outgoing** — rewrites a draft toward a target register (concise, warm, direct, formal) without sanding off your voice. 4. **Brief meeting** — turns a wall of meeting notes into a short brief you can actually use the next morning. ## Where it runs Any image or text content on a page that he user explicitly selects for the extension to translate or explain. ## How to invoke it - Select text and right-click — pick "NeuroDock: translate selection". - Or open the popup and paste a draft into the composer. - Or use the floating button that appears in the supported message composers. Results land in an in-page panel beside the original text. Nothing is auto-applied to your draft — every change is yours to accept or reject. ## Proactive guardrails (new in v0.0.24) Three lightweight checks the extension runs before it asks any LLM to do work for you: - **Hyperfocus check** — if you are deep into the same surface for an unusually long stretch, the extension notices and offers a break, rather than feeding the loop. - **Rumination check** — if you are translating roughly the same message for the fourth or fifth time, the extension says so and suggests stepping away. - **Sycophancy check** — runs on outgoing rewrites; flags drafts where the suggested rewrite would erase too much of your original meaning or voice. All three are advisory, not blocking. You can disable them per check in Settings. The defaults are tuned conservatively. ## Reader preferences Pick how the translator addresses you and how it presents reads: - **Audience framing** — neurotype-aware, neutral-professional, or silent. - **Detail level** — terse one-liner, normal, or expanded. - **Output language** — separate from your interface language, in case you read in one language and write in another. These tune the prompt; they do not change the underlying surfaces. ## Local-first means local-first On install, the extension makes zero network calls. There is no analytics SDK, no error reporter, no telemetry, no remote logging, no crash reporting, no fingerprinting. Out of the box it runs in mock mode and returns clearly-labelled placeholder responses so you can see the surfaces without sending any data anywhere. When you want actual translations, you choose a provider: - **LM Studio or Ollama (local).** The extension talks to a local server on your own machine (or a Tailscale node, or any host you grant permission to). Nothing leaves your device. - **Anthropic, OpenAI, or OpenRouter (cloud).** The extension calls that provider's API directly from your browser, signed with your own API key. NeuroDock has no server in the middle and never sees the traffic. Whenever cloud mode is on, the extension shows a persistent banner in the popup and in every in-page panel. The banner cannot be dismissed without switching back to local or mock mode. You should always know when your text is leaving your machine. ## What is stored, and where - A small profile (your mode, your preferences, your reader settings) in `chrome.storage.local` on this device. - Your API key, if you set one, in `chrome.storage.local` on this device. - Optionally, a local history of recent translations in an extension-scoped IndexedDB database — off by default and capped to a 256-character preview per entry. No remote sync exists. That is the complete list. There is no NeuroDock account, no sign-in, no cloud sync. Uninstalling the extension deletes all of it. The full privacy policy is at https://github.com/tlennon-ie/neurodock/blob/main/PRIVACY.md and the extension-specific surface at https://github.com/tlennon-ie/neurodock/blob/main/packages/extension-browser/PRIVACY.md. ## What it is not NeuroDock is software. It is not therapy, not a medical device, not a diagnostic tool, and not a substitute for clinical care or proper workplace accommodations. It does not treat or remediate any condition. The guardrail layer is documented in `ETHICS.md` in the repository. ## Open source NeuroDock is AGPL-3.0-or-later. Source, manifesto, and contributing guide: - Repository: https://github.com/tlennon-ie/neurodock - Manifesto: https://github.com/tlennon-ie/neurodock/blob/main/MANIFESTO.md - Ethics: https://github.com/tlennon-ie/neurodock/blob/main/ETHICS.md - Issues: https://github.com/tlennon-ie/neurodock/issues