BIOSEEKI · EVIDENCE, NOT GUESSWORK

Read, analyze, write.Every step backed by evidence.

An AI research partner built for individual scientists. Citations lead back to the exact page and paragraph, analyses leave a record you can recompute, and you always make the call.

Download for Mac Create account

Mac Beta · macOS 11+ · Apple siliconWatch the 45-second demo

01 · Capabilities

One place for the everyday work of research.

The everyday research tasks, all inside one project.

GC / MOLECULAR DESIGNIntegrated

GeneCode · Molecular cloning

View sequences and plasmid maps, design primers and cloning plans. Edits the AI proposes wait for your approval.

Open GeneCode on the web
pUC19 2,686 bp EcoRI · BamHI · HindIII lacZα AmpR ori

Deep research

Search, compare the evidence, write it up.

Data analysis

Stats and omics, every run recomputable.

Writing

Drafting, polishing and reviewer replies.

Figures

Laid out and coloured to journal specs.

02 · Verifiable

Every conclusion, traceable to its source.

The most dangerous mistakes in research are the ones that look right.

Back to the source

Citations carry page and paragraph — one click to check.

Citation check-up

Before you submit: DOIs, retractions, versions.

Recomputable analyses

Code, environment and output hashes, all recorded.

Gaps stay gaps

No evidence, no claim: it stays a marked gap.

03 · Knowledge & memory

Sources you can find. A project that remembers.

Put your papers in the knowledge base: it remembers the background and keeps the sources.

Evidence chain · illustrativeSourcesKnowledge-base fragmentsYour draft

Sources

  • PubMed
  • Europe PMC
  • UniProt
  • Ensembl
  • GEO
  • Your PDFs & notes

Knowledge-base fragments

zeb1-ferroptosis.pdfp. 4 · para. 12
emt-review.pdfp. 9 · para. 3
UniProt · P37275ZEB1 · human

Rebuildable index · originals read-only · sha256

Your draft

Mesenchymal-state tumour cells are more sensitive to ferroptosis inducers.[1] p. 4 · para. 12

ZEB1 maintains the mesenchymal phenotype by repressing E-cadherin transcription.[2] p. 9 · para. 3

ZEB1 is a zinc-finger transcription factor.[3] UniProt P37275

⟦Gap: independent replicates after ZEB1 knockdown⟧

Every claim links back to the source. If it can't, it stays out.
P·01 / EVIDENCE

Page and paragraph

Results carry file, page and paragraph.

P·02 / READING

Side-by-side reading

Translation beside the original; citations still point to it.

P·03 / MEMORY

Memory you control

Global and project memory kept apart; view, edit or revoke any time.

04 · A project

From one question to a draft that holds up.

Six steps one project takes in BioSeeki.

  1. 01

    Search the literature

    Journals and preprints listed separately.

  2. 02

    Back to the source

    Key findings bound to their paragraphs.

  3. 03

    Check the citations

    A preprint is now published — cite that.

    You decideWhether to cite the published version

  4. 04

    Design primers

    Three candidate pairs, rechecked in primer3.

    You decideWhich pair to order

  5. 05

    Analyze the data

    Results and plots, with a recomputable run record.

  6. 06

    Draft the text

    Three claims sourced; the unsupported one stays a gap.

High-impact actions need approval

Nothing is written or edited until you approve.

Projects stay separate

Search and memory stay inside a project.

Only open channels for papers

Open and institutional access only. No paywall workarounds.

Your data doesn't train models

Private content is not used to train models.

05 · Real demo

See how it actually works.

45 seconds of the Mac Beta on a real machine. The interface is in Chinese.

BIOSEEKI / PRODUCT DEMO45 SEC · MAC BETA
Edited, with private information hidden. What's shown is not an acceptance test of a complete research task.Download video

06 · Download & get started

Download the app and start with your next research question.

Self sign-up, no invite code, and no model API of your own to set up.

BioSeeki for macOS

Mac Beta
Chip
Apple silicon (M1 / M2 / M3 / M4)
System
macOS 11 or later
Windows
In progress

Intel Macs are not supported yet.

The installer is being signed and notarized; the download opens shortly.

Get the installer by email

Downloading and using it means you accept the Terms and Privacy Policy. Questions: support@bioseeki.com.

Three steps

  1. 01

    Create an account

    Self sign-up, no invite code.

    Create account
  2. 02

    Create a key

    One click in the dashboard; the app connects with it.

  3. 03

    Install the app

    Paste the key, pick a folder, go.

07 · Pricing

Pick a plan and start your research.

Monthly subscriptions, billed in CNY. Each plan includes a fixed credit allowance that expires with the plan. New accounts get 1,000 credits.

Plus

¥69/ month

2,000 credits

Everyday Q&A and light literature search

Subscribe

Pro

Most popular

¥149/ month

5,000 credits

Regular analysis and long-form writing

Subscribe

Heavy

¥299/ month

12,000 credits

Heavy long-running tasks and deep reasoning

Subscribe

08 · FAQ

Things you might want to know.

Didn't find your answer? Write to us

How is BioSeeki different from a general AI chatbot?

BioSeeki works around your project folder: it reads your papers and data, calls research skills and public databases, and keeps sources, run records and outputs in the project. Citations trace back to the original, numbers can be recomputed, and missing evidence is flagged rather than filled in.

Are my files uploaded?

Project files stay on your computer. What goes to the model is only what you type, files you explicitly reference, and the fragments actually read to complete the current task — read on demand, never a whole folder. Private research content is not used to train BioSeeki's models.

Which systems are supported?

The Mac Beta needs macOS 11 or later on Apple silicon (M series). Intel Macs aren't supported yet; a Windows version is in testing.

Do I need my own model API?

No. After signing up, create a key in the dashboard and paste it into the app. Usage is counted in credits; view your allowance and subscriptions in your account.

Will it make up citations or data?

The workflow is built to prevent it: citations must read back to a page and paragraph; before submission you can check DOIs, retractions and versions; and claims without evidence stay out of the text as marked gaps. AI can still make mistakes, so scientific judgement stays with you.

What is GeneCode?

GeneCode is a molecular-biology sequence workbench with feature annotation, primers and restriction sites. Inside BioSeeki it appears as the cloning panel, and sequence edits proposed by the AI are written only after you approve them. It also works on its own in the browser.