Cross-tab synthesis
Open 5–10 papers as tabs and ask 'compare the methodology sections'. Sephir's agent reads each one in sequence and writes a Markdown comparison.
For researchers
Sephir for researchers.
Open every paper as a tab. Sephir's agent reads them as a set — extractPageText across the whole reading list, then a synthesis you can copy. BYOK for unmetered long-context.
What you actually use
Four features you’ll touch every day.
Schema-aware extraction
Give the agent a JSON schema for the data points you want (authors, year, dataset size, accuracy). It returns clean structured output across every tab.
Long-context BYOK
Claude Sonnet handles 200k+ tokens. Paste your Anthropic key — pay Anthropic directly, no markup, no daily cap.
Conversation export
Save the literature-review session as Markdown for your notes app, or JSON for downstream tooling.
A typical day
Five steps, one prompt.
- 01Pin every paper or arXiv tab you want in the comparison.
- 02Open the Sephir sidepanel, paste your Claude API key once.
- 03Ask 'compare these on dataset, model, top-line metric'.
- 04Receive a Markdown table with cells citing back to each source tab.
- 05Export to Markdown, drop into Obsidian / Zotero / your project notes.
Recommended setup
Anthropic Claude Sonnet for long-context literature synthesis. Gemini 2.5 Pro on a BYOK key for cheap quick reads. ChatGPT Plus OAuth for everyday Q&A.
Try the agent layer free.
BYOK from day one, ChatGPT Plus OAuth in one click, no credit card required for the 7-day trial of Pro.