How it works
A searchable library, honest previews, and AI generation grounded in real source material rather than invented from nothing.
The search flow
-
1
You search a topic
Type the assignment topic in your own words. We match against titles, abstracts, keywords and the full text of every published paper.
-
2
We check the library
If we already have relevant work, it appears ranked by relevance. If we don't, we say so plainly and offer to write it.
-
3
You preview for free
Every result shows a real preview — the abstract and opening, plus the section headings. No paywalled mystery boxes.
-
4
You unlock, or you generate
Unlocking an existing paper costs a flat 1,200 tokens, forever. Generating a new one costs by length, and you see the estimate before you commit.
Why grounded generation matters
A plain chatbot writes from memory, which is how fabricated citations happen. We use retrieval-augmented generation instead: before Agentic Writer writes anything, we search our knowledge base of source documents and hand it the most relevant passages.
- Hybrid retrieval. Passages are ranked by both keyword relevance (BM25) and semantic similarity, then the two rankings are fused.
- Inline citation. Agentic Writer cites the supplied passages as [Source N] at the point of use, so you can check every claim.
- No invented references. Where no source covers a point, the model writes from general knowledge and says so rather than manufacturing a citation.
Token pricing by word count
Choose any word count of 500 or more. The cost is 0.750 tokens per word, with a minimum charge of 250 per generation.
| Word count | Estimated cost |
|---|---|
| 800 words | 600 |
| 1,800 words | 1,350 |
| 3,500 words | 2,625 |
| 6,000 words | 4,500 |
This is the amount held when a generation starts, and the exact amount charged — no reconciliation, no surprises. If a run fails, you get the whole hold back.
A word on academic integrity
Everything here is a study aid: a model answer to learn structure and argument from, and a research shortcut to find the relevant literature faster. Submitting generated work as your own is plagiarism at most institutions. Read it, learn from it, verify the sources, and write your own.