Privacy Policy

SeqBench (https://seqbench.com) provides bioinformatics sequence-analysis tools through a website, a JSON REST API, and a Model Context Protocol (MCP) server. This page describes what data is processed when you use any of these surfaces; see the Terms of Service for the rules governing use of the service.

Account-free by default

The core tools, REST API, and MCP server can be used without creating an account or providing credentials. Signing in is optional and unlocks nothing that is otherwise withheld.

Signing in with Google

If you choose to sign in, SeqBench uses Google as the identity provider, through Supabase Auth. Your Google password is never seen by or sent to SeqBench. Google returns your user ID, email address, name, and profile picture URL, which are stored in the SeqBench authentication database so the site can recognise you on your next visit; nothing else about your Google account is requested or received. Signing in also tells Google that you signed in to SeqBench.

You can sign out from the account menu at any time, which deletes the session cookies from your browser. To delete the account record itself, contact us — see below.

Not for clinical or patient data

SeqBench is built for research, educational, and general molecular-biology workflows. It is not designed, validated, or intended for use with Protected Health Information (PHI) or other patient-identifiable clinical data, and must not be used as part of a clinical diagnostic process. See the Terms of Service for details.

Sequence data is not stored

Every tool is a stateless, pure function of its input. Sequences and files you submit — whether pasted into the website, sent to the REST API, or passed as arguments to an MCP tool call — are processed in-memory to compute a result and are not written to a database, logged, or retained afterward. Tools that look up public records (such as accession lookup) send only the public identifier needed to fetch that record from the upstream database.

Most of the workbench does not reach a server at all: editing, restriction digests, the virtual gel, translation, primer design, the cloning simulators, the plate-scale clone verdicts, the folder library, the primer inventory and .dna/GenBank import and export all run in your browser, and a folder you adopt is indexed where it sits — its files are read locally and never uploaded. Ten workbench capabilities do call a server, and each says so on its own panel: deep annotation, plasmid identification, Golden Gate junction fidelity, CRISPR off-target screening, the CRISPR HDR donor, RBS design, construct verification, Sanger readback verification, fetch-by-accession, and the AI assistant described next.

The AI assistant is the one exception

SeqBench has an optional AI assistant — the workbench panel, and the chat on the SeqBench GPT page. It is the one feature that is not a pure function computed by SeqBench: to answer you it forwards your messages, and a projection of the open document, to DeepSeek, the model provider, whose own terms and privacy policy then apply to that text. The projection carries the construct's name, length, topology, feature table and the selected span, and it can be set to a metadata-only mode that withholds the bases entirely. Every number and every sequence the assistant reports comes from a SeqBench tool call rather than from the model, and those calls follow the rules above; but if you would rather no part of a construct leave your browser, do not open the assistant — nothing else in the workbench needs it, and nothing is withheld from you for not using it.

What is logged

To monitor reliability, each API/MCP call records anonymous usage metadata: which tool was called, whether it succeeded, how long it took, the size of the input/output in bytes, and a client identifier string (such as an MCP client or user-agent name). This metadata never includes the sequence, file contents, or any other submitted data. Our hosting and network infrastructure providers may keep standard connection-level logs (such as IP address and request timestamp) for a limited time for security and abuse prevention, consistent with normal web server operation.

Cookies and analytics

SeqBench uses Umami, a privacy-focused analytics service, to understand aggregate site traffic — for example, which pages are visited and how many visitors the site receives. Umami does not use cookies and does not track individuals across other websites, and this analytics traffic never includes any sequence, file, or other data you submit to a tool. SeqBench does not use advertising trackers, and does not sell or share data with advertisers.

The one exception is signing in: if you do, SeqBench sets session cookies that keep you signed in. They are strictly necessary for that purpose, are not readable by JavaScript, are not used for tracking or advertising, and are removed when you sign out.

Third parties

SeqBenchis hosted on Railway and uses Supabase to store the anonymous usage metadata described above and, if you sign in, to run authentication and hold your account record. Signing in involves Google as the identity provider, under Google's own privacy policy. Tools that fetch public sequence records (for example from NCBI) contact that upstream provider directly with only the identifier you supplied. The AI assistant sends what you ask it, and the document projection described above, to DeepSeek as the model provider. Apart from that one feature, no submitted sequence data is shared with, or sold to, any third party, and none of it is ever sold to anyone.

Your rights

If you have an account, or otherwise believe SeqBench holds personal data about you — for example, an email address or standard connection-level server logs — contact us to have it reviewed, exported, or deleted, consistent with applicable data protection law such as the GDPR or CCPA.

Changes

This policy may be updated as SeqBench adds features. Material changes will update the date above.

Contact

Questions about this policy can be sent to seqbench.bio@gmail.com.