Jev in production › Search, ranking and matching
jev-digest
Jev extracts the verbatim passages that answer a query for this MCP tool, cutting input tokens by 67% and tool rounds in half (jev-digest).
67%measured against a baseline, as published by the source
- Use
- Search, ranking and matching
- Industry
- Developer tools
- Form
- Open-source tool
- Stage
- Announced
- Listed
- 2026-09-26
- Found via
- discord
- Repository
- pr0ta9/jev-digest
- Stars
- 0
- Forks
- 1
- Last push
- 2026-09-25
- Language
- Python
- License
- MIT
The README opens with
Turn web pages and local documents into bite-sized evidence, with the original wording and sources intact.
jev-digest searches the web or reads local documents, then uses TypeSafe's Jev model to select the passages relevant to your question. Your agent gets original source text with URLs or file paths attached, ready to reason over and cite.
In one Codex research run: 67% fewer input tokens and half as many tool rounds. See the early results for the full comparison and its limits.
Badge
For the project's own README, linking back here:
Also used for search, ranking and matching
- jegrep $0.01–0.03
- pi-web-search ~2k free uses/month
- Jev Graph Search 81.8%
- Jev RAG 0.444
- oko (bartlomein) 13–44% fewer tokens
- jevsearch 83%
- GStars 1500
- askgrep $0.0049