The upstream data changed shape: from the 2026 edition the warning status is carried inline as a field rather than published as a separate list, and the tool scans both. That matters because a journal flagged under only one scheme would otherwise come back clean. The trend tool is the other reason to use this rather than a single lookup — a partition that has slipped across editions tells you something a current-year figure hides. Note the data is a local snapshot: it is as current as your last sync.
A journal-metrics lookup built on the ShowJCR data set: JCR impact factors, the Chinese Academy of Sciences partition tables, the international journal warning lists, the CCF recommended venue catalogue, and the computing-field quality grading. The data is synced once into a local SQLite database and queried from there.
- A journal looked up by name, returning impact factor, partition and warning status together rather than from three separate sources (search_journal).
- How a journal's partition has moved across editions, which is the question a single year's figure cannot answer (get_partition_trends).
- The international journal warning list searched by keyword, scanning both the traditional warning tables and the 2026 edition's embedded warning marker, so old and new sources are covered in one call (check_warning_journals).
- Several journals compared side by side in one call rather than looked up one at a time (compare_journals).
- Database information and statistics exposed as a resource, so you can see which editions the local copy actually holds.
- A journal-analysis prompt template shipped with the server.
Python 3.10 or newer — a hard requirement of the MCP SDK — plus SQLite, which ships with Python. Install the dependencies, then run the data sync once before first use; it downloads and imports the tables locally. After that the server runs offline against that database. No account and no key.
One command — claude mcp add -s user jcr-partition -- /path/to/python /path/to/jcr_mcp_server.py