Skip to main content
Bower is your built-in AI assistant. It has full access to your workspace — it can read, search, create, and organise your notes, collections, and attachments. It can also search the web for scientific sources. Every answer Bower gives includes inline citations — numbered references back to the specific notes, attachments, or web sources it used. You always know where the information came from.

​
What Bower can do

Ask Bower in plain language: “Show me pH readings from last month” or “Find my notes on sample batch 14.” Bower searches across everything in your workspace — voice transcripts, photo extractions, imported files — and returns relevant results with links to the source notes.
Bower can search the web — and the open scholarly record — for external sources. Ask “What’s the latest research on coral alkalinity tolerance?” or “What did Jane Smith publish in 2025?” and Bower will find relevant sources and cite them alongside your workspace data.Out of the box, Bower searches eight open sources for you, with no setup:
  • The open web via Google, including Google Scholar, journal sites, and institutional repositories.
  • OpenAlex — the open catalogue of ~250M scholarly works. Topical search + fetch by DOI.
  • Semantic Scholar — ~200M papers with AI-generated TLDRs, citation counts, and open-access PDF links. Bower also uses Semantic Scholar’s citation graph to walk both the references and the forward citations of any paper you’re interested in.
  • Crossref — the canonical DOI registry, used as a fallback when other indexes are sparse and as a cross-check on the canonical publisher record.
  • Unpaywall — given a DOI, Bower finds the best-known open-access copy so you can read the full text without leaving Bower.
  • arXiv — preprints across CS, ML, physics, math, statistics, biology, and economics. Bower fetches the latest preprint version when one is available.
  • Europe PMC — biomedical and life-sciences literature (PubMed + PMC + bioRxiv / medRxiv preprints) with full-text URLs.
  • ORCID — public publication histories. Bower can list a researcher’s works by their ORCID iD.
Save your ORCID iD on your profile (Settings → Profile → Research identity) and Bower will automatically pull from your publication list when you ask things like “what have I published recently on coral alkalinity?” — no need to share the iD each turn. We only ever read your public ORCID profile. Add yours now.
Give Bower one or more papers you already know are relevant, and it can walk the citation graph for you — pulling in both the foundations (papers cited by your seeds) and the follow-on work (papers that cite your seeds). Bower dedupes and ranks the results so the most-cited and most-co-cited papers float to the top, and resolves open-access PDF links automatically.Ask Bower things like:
  • “Starting from DOI 10.1038/s41586-023-XXXXX, find the most relevant follow-on work.”
  • “I’m reading these three papers. What else should I be reading?”
This is dramatically more effective than running flat searches when you already have one or two papers on the topic — citation-graph proximity is a much stronger relevance signal than keyword overlap.
Select one or more notes and ask Bower to summarise. Useful when you’ve captured a week of observations and want the key points before a meeting or write-up.
Bower can pull related notes side by side. “Compare my alkalinity readings from this week and last week” works as a natural language query.
Bower can create new notes, protocols, and collections on your behalf. “Create a protocol for the PCR workflow we discussed” or “Make a new collection called February Field Samples” — Bower handles the creation and confirms what it did.
Bower can move notes between collections, link related entities, add tags, and update titles. “Move all my alkalinity notes into the Field Samples collection” saves you the manual work.
Bower can trigger text extraction on images, transcribe audio files, and read the content of PDFs and documents in your workspace.
Given a set of notes and a protocol, Bower can draft a structured execution plan. Useful for turning a series of captured observations into a formal experimental record.
When a captured value falls outside the expected range for a protocol field, Bower surfaces it after capture — not during. It explains why it flagged the value and asks what you’d like to do.
Workspace owners and admins can ask Bower to generate an invitation link: “Create an invite link for the lab” — Bower generates a shareable link you can send to colleagues.
Ask Bower to update your display name or save your ORCID iD without leaving the chat. “Save my ORCID 0000-0001-2345-6789” or “Set my display name to Dr Jane Smith” both work. Bower verifies the ORCID iD against ORCID’s public API before saving — exactly the same check the Profile tab runs — and replies with the canonical name it found, so you know the iD belongs to the right person. To remove a saved iD, ask “clear my ORCID iD”.

​
Citations — always showing the source

Bower never gives you an answer without telling you where it came from. When Bower references a source, it places a numbered citation marker like [1] directly after the relevant claim. You can click the marker to see the source details. At the bottom of each response, a collapsible Sources section lists every source Bower used:
  • Workspace sources — links to the specific notes, protocols, or attachments Bower referenced. Click through to read the originals.
  • Web sources — links to external articles, papers, or documentation Bower found online. Favicons and page titles are shown for easy identification.
This means you can always verify Bower’s answers. Citations appear for both workspace searches and web searches, so every factual claim is traceable.

​
How to use Bower

Bower is available in the right-hand panel of the desktop app. Type your question or instruction in natural language. No special syntax required. You can also use context chips — select specific notes or collections before asking Bower a question, and it will focus its response on those items.

​
Attaching files

You can give Bower a file to work with directly in the chat input. Three ways, all equivalent:
  • The paperclip button in the chat input opens a file picker.
  • Drag a file from your desktop or file manager onto the chat input and drop it.
  • Paste a file — a screenshot you’ve just taken with your clipboard, for example.
Attached files appear as chips above the input before you send. Remove one with the × on its chip. They’re attached to that message only, so each question can carry its own files. Bower accepts documents, images, spreadsheets, PDFs and research code — including scripts like .R, .Rmd, .jl and .do that many tools reject because the operating system doesn’t give them a recognised file type. If a file is too large or isn’t a type Bower can read, you’ll get a message saying which file and why, and the rest of your selection still attaches.
Dropping a photo of a gel, plate or instrument display straight into the chat is the quickest way to ask Bower about it — no need to save it to disk first.
Pasting a table copied out of a document pastes it as text, not as an attachment. To have Bower read a table, attach the file itself or a photo of it.

​
Watching Bower work — the live activity stream

While Bower is working on a question, the response area shows a live activity line — short, plain-language descriptions of what Bower is doing right now: Searching notes…, Reading a Drive file…, Looking up your publications on ORCID…. Each step appears as it happens, replacing the older rotating “Thinking…” placeholder. Once Bower finishes, the live line collapses and the assistant’s full reply appears. The activity stream is purely informational — there’s nothing to click — but it’s useful when a question takes longer than expected: at a glance you can see whether Bower is still actively searching or stuck waiting on an external service.

​
Voice dictation

Instead of typing, you can tap the microphone button in the chat input to dictate your message. Speak naturally — your words appear as text in the input field in real time. Tap the stop button when you’re done, review the text, and send. Voice dictation works on both desktop and mobile:
  • Chrome, Safari, and Edge — text streams live as you speak. You can see and review your words before sending.
  • Firefox — your voice is recorded first, then transcribed when you stop. Text appears after a short processing delay.
If your browser doesn’t support voice input, the microphone button won’t appear.
Voice dictation is great for long scientific questions — speak your query instead of typing complex terminology.

​
Live conversation

For a continuous back-and-forth instead of one-off dictation, tap the sparkle button in an empty Bower chat. That launches a live agent session — real-time voice and (optionally) video, with hands-free snapshot and clip capture. Available on mobile and desktop.

​
Rating Bower’s responses

Every Bower response has thumbs up and thumbs down buttons. Rating responses helps us improve Bower for scientific workflows — both positive and negative feedback is valuable.

​
What Bower won’t do

  • Bower won’t act on your data without permission. It suggests; you confirm.
  • Bower won’t present AI-extracted values as verified fact. Confidence indicators are shown where relevant.
  • Bower won’t interrupt a capture in progress. It waits until you’ve stopped recording before offering observations.
Bower and search are different tools for different moments. Use search when you know what you’re looking for. Use Bower when you’re trying to make sense of what you’ve captured, or when you need to take action.