ScholarFetch v0.2.1

ScholarFetch v0.2.1 turns the project into a more stateful academic research environment for both humans and agents.

Highlights

• Added a stateful MCP research scratchpad through named in-memory reading-list collections
• Added MCP curation/export tools:
  • scholarfetch_saved_add
  • scholarfetch_saved_list
  • scholarfetch_saved_remove
  • scholarfetch_saved_clear
  • scholarfetch_saved_export
• Added export modes for downstream synthesis:
  • citations
  • abstracts
  • bib
  • fulltext
• Improved MCP tool descriptions so agents can understand valid parameters and workflows more easily
• Added agent-facing documentation:
  • AGENTS.md
  • SKILL.md
  • SKILLS.md

CLI Improvements

• Tree-based research traversal picker
• References treated as paper-like nodes
• Author expansion from selected papers, including multi-author exploration
• Saved-paper workflow inside the interactive session
• Better full-text and reference navigation
• Clearer availability hints for abstract and full text
• Better progress feedback during longer retrieval steps

MCP / Agent Workflow

ScholarFetch can now support a stronger agent loop:

1. search for candidate papers
2. inspect abstracts and full text
3. expand references and authors
4. keep a curated saved set in a named collection
5. export a compact research corpus for synthesis

This makes ScholarFetch more useful as a research workspace rather than a one-shot search endpoint.

Public MCP Endpoint

• https://laibniz-scholarfetch-web.hf.space/mcp/

Registry

• MCP Registry name: io.github.laibniz/scholarfetch

Notes

• This release includes updated remote MCP registry metadata (0.2.1)
• No secrets or local environment files are included in the repository