alphaxiv-paper-lookup

AlphaXiv Paper Lookup

Workflow

Step 1: Extract the Paper ID

Parse the paper ID from whatever the user provides:

Input	Paper ID
https://arxiv.org/abs/2401.12345	2401.12345
https://arxiv.org/pdf/2401.12345	2401.12345
https://alphaxiv.org/overview/2401.12345	2401.12345
2401.12345v2	2401.12345v2
2401.12345	2401.12345

Step 2: Resolve the Paper

curl -s "https://api.alphaxiv.org/papers/v3/{PAPER_ID}"

From the JSON response, extract:

• versionId — UUID needed for Step 3
• title — paper title
• authors — author list

If this returns 404, the paper has not been indexed on alphaxiv yet — tell the user.

Step 3: Fetch the AI Overview

curl -s "https://api.alphaxiv.org/papers/v3/{VERSION_ID}/overview/en"

The response contains:

• intermediateReport — machine-readable structured text; best for LLM consumption
• overview — full markdown blog post; human-readable
• summary — structured fields: summary, originalProblem, solution, keyInsights, results
• citations — cited papers with titles and justifications

Prefer intermediateReport when available. Fall back to summary fields if intermediateReport is null.

Step 4: Present the Overview

Synthesize the fetched content into a clear, readable response. Do not dump raw JSON. Structure the output with:

• Title and authors (from Step 2 metadata if available)
• Problem being solved
• Approach / solution
• Key findings and results
• Notable citations (if relevant to the user's question)

Error Handling

• 404 on Step 2: Paper not indexed on alphaxiv. Tell the user it is not available yet.
• 404 on Step 3: Overview not yet generated for this version.
• intermediateReport is null: Use summary and overview fields instead.

Notes

• No authentication required. These are public endpoints.
• If bash_tool is unavailable, use web_fetch on the same URLs — the responses are identical.
• Replace en with a language code for translated overviews: fr, de, es, zh, ja, ar, hi, pt.
• Version suffixes (e.g. v2) are preserved and passed through as-is.