Asta Skill

Asta MCP — Academic Paper Search

Asta is Ai2's Scientific Corpus Tool, exposing the Semantic Scholar academic graph over MCP (streamable HTTP transport). This skill tells agents which Asta tool to call for which intent, and how to compose them into useful workflows.

• MCP endpoint: https://asta-tools.allen.ai/mcp/v1
• Auth: x-api-key header (request key at <https://share.hsforms.com/1L4hUh20oT3mu8iXJQMV77w3ioxm>)
• Transport: streamable HTTP

Prerequisite Check

Before invoking any tool, verify the Asta MCP server is registered in the host agent. Tool names will be prefixed by the MCP server name chosen at install time (commonly asta__<tool> or mcp__asta__<tool>).

If no Asta tools are visible, do not make raw HTTP calls or invent results. Tell the user to register https://asta-tools.allen.ai/mcp/v1 as a streamable HTTP MCP server with an x-api-key header, then restart/reload the host. Minimal setup hints:

• Codex CLI: add [mcp_servers.asta] url = "https://asta-tools.allen.ai/mcp/v1" and env_http_headers = { "x-api-key" = "ASTA_API_KEY" } to ~/.codex/config.toml.
• Claude Code: run claude mcp add -t http -s user asta https://asta-tools.allen.ai/mcp/v1 -H "x-api-key: $ASTA_API_KEY".
• Generic MCP clients: configure server URL https://asta-tools.allen.ai/mcp/v1 with header { "x-api-key": "<YOUR_API_KEY>" }.

Tool Map — Intent → Asta Tool

Most search/citation tools accept `publication_date_range` (format YYYY-MM-DD:YYYY-MM-DD; year shorthand like "2021:", ":2015-01", "2015:2020" is also accepted), `venues`, and `fields` for field selection — pass them whenever the user's intent constrains scope (e.g., "recent", "since 2022", "at NeurIPS"). venues matches Semantic Scholar's exact venue strings (comma-separated, e.g. "Nature,N. Engl. J. Med."); a casual name like "NeurIPS" may not match, so fall back to a date/keyword filter when a venue lookup returns empty.

Per-tool parameter exceptions (verified against the live server — getting these wrong yields a malformed or silently-ignored argument):

• get_author_papers names its field-selection param `paper_fields`, not fields (passing fields= is silently ignored — you get titles only), and accepts a publication_date_range but no venues filter. Its date filter is applied after paging and can drop valid in-range papers (a well-known author's 2024:2024 returned []), so for reliable results narrow by topic/venue and date client-side.
• get_citations accepts publication_date_range but not venues.
• snippet_search accepts neither fields nor publication_date_range. Instead it has: `inserted_before` (date filter, YYYY-MM-DD/YYYY-MM/YYYY), `paper_ids` (comma-separated list of ≤100 IDs to restrict snippets to specific papers), and venues.

⚠️ fields parameter — avoid context blowups

get_paper / get_paper_batch accept a fields string. Never request `citations` or `references` via fields — a single highly-cited paper (e.g. Attention Is All You Need) returns 200k+ characters and will overflow the agent's context window. Use the dedicated get_citations tool for forward citations (it paginates). Asta does not provide a dedicated get_references tool — to retrieve a paper's reference list, use get_paper with fields=references only for papers you know have a small reference list (typically < 100).

Watch row counts too, not just per-row size: default limits are large — get_author_papers returns up to 1000, get_citations 100, search_papers_by_relevance 50, and snippet_search 20 (each snippet is ~500 words, making it the heaviest tool per row). Pass an explicit small limit — e.g. 20–50 for paper/citation lists, ~5–10 for snippets — unless the user asked for the full list.

Use task-specific field presets:

Metadata lookup:

title,year,authors,venue,tldr,url,abstract

Search/ranking/results tables:

title,year,authors,venue,tldr,url,abstract,citationCount,influentialCitationCount

DOI/export handoff:

title,year,authors,venue,tldr,url,externalIds

Add journal, publicationDate, fieldsOfStudy, isOpenAccess only when needed. If pass-through fields such as citationCount are absent in a future response, degrade gracefully: sort by relevance/recency and omit citation-based claims.

Example call (topic search, ranked by citations, DOI exposed for downstream fetch):

search_papers_by_relevance(
  keyword="mixture of experts routing",
  publication_date_range="2023:",
  fields="title,year,authors,venue,tldr,url,externalIds,citationCount",
  limit=20,
)

Retrieving DOI / external IDs (undocumented but supported)

Asta's official fields list does not include externalIds, but the field is transparently passed through to the underlying Semantic Scholar API and works in practice. Add externalIds to fields to retrieve DOI, PubMed, PubMedCentral, ArXiv, MAG, DBLP, CorpusId. The same pass-through applies to `citationCount` and `influentialCitationCount` (also absent from the official list but verified to return) — request them when ranking results by citations. Caveats:

• Not all papers have a DOI — pure arXiv preprints often only return ArXiv + CorpusId.
• get_paper("DOI:...") lookup is not 100% reliable; some valid DOIs return not found. Prefer searching by title first, then reading externalIds off the result.
• Since this is undocumented, treat it as best-effort and degrade gracefully if a future Asta release drops it.

Workflow Patterns

Pattern 1 — Topic Discovery

1. search_papers_by_relevance(keyword, publication_date_range="<current_year-5>:", venues=?, fields="title,year,authors,venue,tldr,url,abstract,citationCount,influentialCitationCount", limit=20) → initial hits (compute the lower bound from today's date — e.g., in 2026 pass publication_date_range="2021:"; adjust or drop the filter if the user asks for older work)
2. Rank/present top N by citationCount + recency
3. Offer follow-ups: get_citations on the most influential, or snippet_search for specific claims

Pattern 2 — Seed-Paper Expansion

1. get_paper(DOI|arXiv|...) → verify seed
2. get_citations(paperId) → forward expansion
3. Optionally search_papers_by_relevance with seed title terms for sideways discovery
4. Deduplicate by paperId before presenting

Pattern 3 — Author Deep-Dive