SciScrape Agent · Evidence-linked AI
SciScrape Agent for evidence-based research
An AI research system designed for scholarly rigor. Search literature, interrogate lab spreadsheets, and audit manuscript drafts — with every claim linked to a verifiable passage or DOI.
Four research profiles, one rigorous workflow
SciScrape Agent adapts its prompts, tool use, and verification methods to the specific task at hand:
Literature & Evidence Review
Evaluates scholarly literature retrieved from six databases, extracts supporting and contrasting quotes, and flags claims supported only by abstracts.
- ✓Distinguishes full-text passages from abstract-only claims
- ✓Identifies conflicting conclusions across multiple papers
- ✓DOI or source links and bibliographic provenance where available
Project Workspace Investigation
Acts as an autonomous pair researcher within your project workspace, combining uploaded PDFs, spreadsheets (XLSX/CSV), notes, and external papers.
- ✓Compares experimental results in spreadsheets with published literature
- ✓Identifies missing control experiments, gaps in evidence, and unverified assumptions
- ✓Generates structured project notes with precise document and page citations
Draft Audit & Preflight
Audits thesis chapters and manuscript drafts before supervisor or journal submission, checking references against the literature.
- ✓Highlights questionable claims, over-generalized conclusions, and small sample sizes
- ✓Verifies in-text citations against the actual referenced texts
- ✓Performs revision checks comparing v1 to v2 to verify revisions were made
Peer Review Assistant
Assists invited peer reviewers by analyzing manuscripts against reporting standards and journal checklists in a private workspace.
- ✓Generates private review notes accessible only to the reviewer
- ✓Checks adherence to methodological and statistical reporting guidelines
- ✓Drafts constructive, respectful reviewer feedback for the author
Execution scopes: Quick and Deep
Choose the right depth for your question. We never claim one mode holds “more truth” — they differ in reasoning steps, literature queries, and computational cost:
Quick mode
A bounded, focused investigation focused on a specific hypothesis, document comparison, or quick claim verification.
Best for: Quick fact-checking, single paper comparison, quick summary notes
Deep mode
Multi-turn autonomous investigation that searches the literature, evaluates conflicting findings, and synthesizes multi-source evidence.
Best for: Comprehensive literature gap analysis, full chapter review, methodology validation
Human agency and source transparency
SciScrape Agent is built around verifiable evidence, not black-box automation:
Human in the loop
SciScrape Agent proposes findings, notes, and actions. Nothing is automatically applied or submitted without explicit human review and acceptance.
Source transparency
Every statement carries a clickable source anchor linking to the exact document, spreadsheet sheet, page, or verified DOI.
Coverage awareness
The agent explicitly reports its reading scope: whether it examined full text, parsed tables, or only reviewed the abstract.
Scientific limitations
Language models can misinterpret nuance or miss critical context. AI interpretations must always be verified by the researcher before publication or submission.
Ready to investigate your research?
Start with a free account, or unlock 90 days of project investigations with the Research Pass.
A practical research example
Compare enzyme treatments in two papers with your own dough measurements. Ask which differences in dose and method prevent a direct comparison.
- Define your question and search for relevant literature.
- Save useful sources in a project and add documents you can share.
- Select the AI context and ask the agent to compare findings.
- Inspect cited passages and limitations before using the findings in your draft.
Frequently asked questions
Does every paper include a PDF?
No. Full text may be restricted. You can add a legally obtained PDF to your project.
Does the agent replace scientific judgement?
No. Check cited evidence and assumptions. Missing evidence and AI mistakes remain possible. Run costs are shown before you start.