Six tools that keep the reading in your hands and the bookkeeping out of them.
IntermediateRequired tools from Free – $84/mo when every seat is on the cheapest monthly tier.
Vendor annual plans, usage billing, and team seats can change the real bill. Treat this as a floor, not a quote.
Test the boundary. Paste the four lines, ask for papers that fit and papers that fail one line, and note which suggestions drift outside the years or population. Two reports a month on the free tier, so spend them on the boundary rather than on browsing.
Verify before you save. Check title, year, venue, and the reference list of every candidate a graph or a summary handed you. Retracted or preprint-only work shows up here first.
Where the clusters live. One entry per opened paper with three lines: claim, measured outcome, open question. Collections carry the clusters, tags carry the anchors. Candidates you have not opened stay out.
Two seeds you have read, one graph, and a triage pass. Each neighbour gets a seed relation and a reason to open it, or it is dropped. Five graphs a month on the free tier keeps the seeds deliberate.
Ask the yes-or-no version of your question and read the paper pairs that split. Take the pairs, leave the aggregate meter. The design difference is what you write down.
Compress cluster notes into candidate headings once the argument already exists. Paste notes, get headings, delete the ones without papers. Never the source of a paper or a finding.
This stack only earns its keep when you actually follow the run order.