The tools qualitative research actually needs
What NVivo and MAXQDA do, in your browser. The AI speeds you up; the decision stays yours.
From text to code
Documents stay immutable; every excerpt is anchored to its exact position.
Transcripts and field notes
Word (.docx), PDF, plain text, subtitles (.vtt/.srt) and spreadsheets. Interviewer and participant turns are separated; timecodes are preserved.
Focus group mapping
In a five-speaker recording you map “S1 → Aisha” once; every turn with that label attaches to the right participant and enters the matrix.
Open-ended survey answers
Pick open-ended questions from a YouReply Survey: each respondent becomes a participant, each answer a document. Demographic questions map to participant attributes.
Code by highlighting
Select text, assign a code. Overlapping highlights render correctly; an excerpt can carry several codes, and co-occurrence analysis comes from that.
Hierarchical codebook
Codes, sub-codes, definitions, inclusion and exclusion rules. Definitions are not only for inter-coder agreement — in deductive coding they are the text the model reads.
Merge, split, move — undo
Every structural operation tells you how many codings it will affect, is recorded with your rationale, and the last one is undoable in a single step.
An assistant that suggests, never decides
The AI does not finalise the analysis. It suggests codes, themes and relations; accepting, editing and rejecting belong to the researcher.
Code suggestions
With a codebook it assigns only existing codes; without one it proposes new ones. Every suggestion is anchored to a verbatim excerpt — an unanchored suggestion cannot be accepted, and its count is reported to you.
Codebook consolidation
Groups codes that describe the same phenomenon. No group merges on its own; you approve them one at a time.
Theme synthesis
Groups your codes into themes and drafts a narrative for each. Themes arrive as CANDIDATES and count for nothing until you activate them.
Pattern contrast
Similarity, difference, contradiction and outlier. An outlier is not noise — it is often the most interesting finding. Every finding is shown with the excerpts it rests on.
From matrix to manuscript
Every number is clickable; real excerpts sit underneath it.
Theme × participant matrix
Who said what. Each cell carries two numbers — how many people and how many excerpts — because “half the participants” and “half the excerpts” are not the same claim.
Code × demographics matrix
Compare by attributes you define: seniority, age group, department. Participants with a missing value appear in their own column rather than being hidden.
Analytic and reflexive memos
Attach them to a document, a code, a theme or a single excerpt. This is where your methods and reflexivity sections come from.
Word and Excel reports
Codebook, quote table, theme narratives and matrices. Quote text comes from the database — the AI never writes a quote into the report.
Why you can defend these findings
A number you cannot defend is useless in an academic study.
Every finding is anchored
The chain theme → code → excerpt → character range in the document → participant is never broken. You can always show where a quote came from.
The AI never finalises alone
An unaccepted suggestion never enters a matrix, a count or a report. How much of the coding started as an AI suggestion is recorded and can be declared in your paper.
Fabricated quotes are structurally impossible
The model does not write quotes into reports; it cites numbered excerpt ids and the text is resolved from the database. An unresolvable citation is removed together with the sentence containing it.
You can switch it off
Turn AI off in the study settings and no text from that study leaves for the analysis service. If your ethics board requires it, this is a checkbox.
Code your first interview today
The free plan carries a pilot study from start to finish. No credit card required.