InTree
The search engine that reads the papers for you. Ask about any drug, gene, disease or claim — across 240M+ papers — and get an interactive evidence map, a clear verdict, or a step-by-step mechanism check.
Every answer shows the exact sentences from the papers that prove it — so you spend your time on science, not on screening PDFs.
Answering one biomedical question usually means reading 50 papers.
Search engines give you a reading list. InTree reads the papers for you — 240M+ of them — and turns what they actually say into a map you can check, sentence by sentence. Here's what that looks like with a real question.
Ask about one thing — in one word
Say you want the full picture on metformin — the world's most-prescribed diabetes drug. What does it treat? What does it cause? How does it work?
You type one word. Before reading anything, InTree shows you its plan: which angles it will read the literature through — drugs, diseases, biomarkers, pathways, outcomes. It explains its choice in one line, and you can change any of it. Then it reads: scanning the knowledge base, cross-referencing the literature, synthesizing — about a minute the first time, instant when cached.
A minute later: a map of what science knows
97 connected things — conditions it treats, side effects, the genes behind it, related drugs — joined by 381 evidence-backed links. Below is the actual InTree workspace with that result loaded, tools and all. Click any circle for its toolbox, click any line to read the papers behind it, try the filters — everything is real and free.
Follow the green TREATS edges: type 2 diabetes, prediabetes, PCOS, gestational diabetes, fatty liver disease.
The map keeps CAUSES separate: gastrointestinal adverse events, diarrhea, vitamin B12 deficiency.
The gene cluster: in via the OCT1 transporter, inhibits complex I (NDUFA10 †), activates AMPK. † = Open Targets, * = curated database.
Cancers, cognitive decline, the microbiome — amber and dashed-blue edges mark emerging and preliminary links.
Every line opens the actual papers
This is the part search engines can't do. Click the line between metformin and type 2 diabetes and InTree shows you the exact sentences from the papers that back it — with the author, year, journal, study type, citation count and a one-click DOI link.
“Metformin is the first-line drug to treat type 2 diabetes mellitus.” — Valle et al., 2022
Sort by citations, influence, study strength or recency. Copy a citation in APA, Vancouver or BibTeX. No black box — you can always see why the map says what it says.
Need a straight answer? Ask for a verdict
“Does metformin actually treat diabetes?” InTree reads each paper's claim and takes a side: supports (a measured result), claimed (proposed, never tested), contradicts, or indirect. Then it weighs them — a clinical trial counts for more than a case report.
Testing a hypothesis? Check it hop by hop
Mechanisms are chains: A activates B, which activates C. Type the chain — EGFR → PI3K → AKT, a famous cancer signalling route — and InTree verifies each hop separately: its direction (activates ⊣ inhibits), its evidence count, its strongest study.
- • A chain is graded by its weakest link, not its average.
- • This one matches 4 expert-curated pathway databases — it says so.
- • Typed “AKT”? InTree resolves it to the canonical gene AKT1 and pools every alias.
- • Thin hops get a generated molecular mechanism — honestly labelled as a hypothesis.
When you're ready, it goes deeper
- Pathways & mechanisms — an on-demand engine traces multi-step routes between entities, separated into honest tiers: established pathways vs hypotheses vs mere associations.
- Research gaps — real pairs with only a handful of papers: under-studied links that might be your next project.
- Clinical trials — 2,098 registered trials indexed for metformin alone, with phases and live status.
- A graph you build on — expand any node, add your own, verify and trace in place; your findings merge into the same map, and the link you share reproduces it exactly.
What would you ask?
Every example below runs free, without an account. Ask in plain English — InTree parses it into the right mode — or use the search box above.
Why researchers trust what the map says
Every tool on the graph, explained
The graph is a workspace, not a picture. Here is the complete toolset — grouped the way you'll find it in the app.
Opens the sidebar toolbox for it: Expand / Verify / Trace / Evidence — and InTree quietly identifies any still-generic edges around it.
Opens the evidence drawer AND runs a direction check against actual paper sentences — 'Direction confirmed by N papers' or a warning that papers suggest the reverse.
Grows the graph in place — that node's neighbourhood merges into your current map, no page reload.
Select several nodes and a 'Trace pathway →' bar appears — send the selection straight into a hop-by-hop trace.
Grey hairlines are co-mention leads. One click and InTree types the relationship — direction, effect, evidence — and the label sticks.
Pan, zoom, and rotate the board (labels stay upright). Node positions you drag are remembered.
Jump-center any seed with number keys or the seed picker — multi-seed graphs stay navigable.
Instead of shrinking into a hairball, the graph holds a readable zoom and gives you a clickable minimap.
Until a relationship's type and direction are settled, the edge renders neutral and undirected — the graph never shows an arrow that could still change.
Now run your own
240M+ papers, mapped. Ask about the entity you actually work on — the first searches are free.