Related-work triage
See which neighbors actually connect to the target, not just which titles share words.
Graph-grounded research
Leave a paper with cited directions, not a chat transcript.
Retrieve related work, build a target-centered graph, and read four graph-grounded findings. Model-generated drafts — not peer review.
Research workspace
Find a paper, then generate cited directions.
Searches OpenAlex. Does not use a generate run. Pick a hit to fill the form.
OpenAlex liveTitle, DOI, arXiv, or PMID. Find does not spend a generate run.
Cited edgesFindings that invent triples are dropped before you see them.
Three runs / hourFree cap is per IP. Shared links and Find stay unlimited.
How it works
From a paper to inspectable findings
01
Paper
Find a work in OpenAlex or paste a title and abstract. Find does not use the generate cap.
02
Multi-query retrieval
Run profiles the target, writes several queries, ranks about twelve bases, and hops their top references.
03
Evidence graph
Five relations: addresses, uses, improves, evaluated_on, limited_by. Up to four graph-derived contexts.
04
Research directions
Four ranked findings first, each citing graph edges. Extra candidates stay in a drawer.
Try a 30-second run
Watch a SciMON-style result, then click in.
This is a fixture, not a live generate. No quota. The graph grows, you pick a finding, evidence lights up. Then start from your own paper.
Silent loop
Ground novelty search in typed graph walks, not raw abstract overlap
Idea generators that retrieve by keyword still let the model wander off the literature.
Playable fixture
SciMON
Method. Extract addresses/uses/improves edges from retrieved work, then ask the model to walk those edges when proposing a follow-up.
Find
Look a paper up in OpenAlex, or paste a title and abstract.
Graph
Related work becomes typed edges around the target.
Evidence
Each finding cites edges that already exist.
Findings
Four ranked directions first. The rest stay in a drawer.
Use cases
Where it helps
Follow-up directions
Read four cited drafts you can argue with, then open the sources they used.
Lab-meeting preparation
Share a run link. The graph and findings travel together for 30 days as a guest, or stay on an account.
Why conscRAG
How we’re different
There are a lot of AI research tools. This one starts from a paper and stays attached to a graph you can inspect.
Cited graph edges
Every finding must cite two to four evidence triples from that run. Invented edges are dropped.
Inspectable retrieval
Several OpenAlex queries, ranked bases, and reference hops — not a single title search.
Honest limits
OpenAlex only. Optional open PDFs. Three generate runs per hour. Not peer review.
More than a chat
A graph, a source list, and four ranked dossiers you can compare and share.
FAQ
Common questions
What is conscRAG?
conscRAG is a research aid operated by The Consc Company in Ontario, Canada. You start from a title and abstract (or find a work in OpenAlex). It retrieves related work with several queries, builds graph-derived contexts, and returns four ranked findings plus more candidates. Those findings are model-generated drafts, not peer-reviewed papers.
What can I type in Find a paper?
One box accepts a title fragment, keywords, a DOI, an arXiv id or URL, or a PMID (as pmid: or a PubMed URL). Press Find or Enter to query OpenAlex and show up to eight hits. Click a hit to fill title, DOI, and abstract. Find does not count against the three generate runs per hour on Free.
Does Run use the same quota as Find a paper?
No. Free use is limited to three generate runs per hour per IP. Searching OpenAlex from Find a paper does not spend that cap. Opening an old share link also does not.
Do you save what I run?
If you continue with Google, the run is saved to your account. Guests get a share link that lasts 30 days. The browser also keeps the last run in localStorage. Find-a-paper queries are not saved as runs.