A clearer toolset for connected AI
The Connect page lists every MCP tool, and the tools now read long papers in sections, find papers by title, page long lists and explain empty searches.
AI tools connected to PaperVantage through MCP, such as Claude, get a clearer and more capable set of tools in this release.
- See what your AI can do. The Connect page now ends with a list of every tool the server offers, grouped by what they're for, and marks which ones only read and which ones change your projects or library. The list comes straight from the server, so it always matches what your client sees.
- Long papers, a section at a time. Reading a paper's full text now returns it in windows, with a marker saying where to continue, so a long paper no longer arrives as one very large answer.
- Find a paper by its title. Paper search has a title mode for looking up one specific paper, for example from a bibliography. It tolerates differences in case, punctuation and subtitles, and if the paper isn't in the library yet it says which paper Semantic Scholar matched the title to.
- Long project lists come in pages. Listing your projects returns them a page at a time, with a total.
- Empty searches explain themselves. When a passage search finds nothing, it says whether the paper hasn't been indexed yet, can only be matched on exact words so far, or simply has no matching passage, so your AI knows whether to rephrase or read the paper instead.
If you've written your own scripts or prompts against these tools, note that paper search, project listing and full-text reading now return their results in a slightly different shape, with the new fields described in each tool's description.