Does a paper reproduce? Read-only lookup of OpenRepro's re-run reports, claims and evidence.
Scored across 6 tools
Most tools have clearly distinct purposes, but check_paper and get_report both retrieve information about one paper and differ mainly in depth. The descriptions do clarify summary versus full public record, so confusion is limited.
All names use consistent snake_case and are mostly verb-led (check_paper, get_report, request_rerun, search_reports, send_feedback). how_to_verify is a minor deviation in verb style but remains clear and consistent with the overall naming pattern.
Six tools is well-scoped for a public ledger server, with each tool covering a distinct workflow: search, lookup, full report, verification, rerun request, and feedback. There is no evident bloat or missing core operation.
The tool set covers discovery, paper-level summary and detail, verification instructions, rerun requests for missing papers, and optional feedback. Evidence links and downloadable bundle guidance are included through the report and verification tools, leaving no obvious dead ends.
Look up one paper in the OpenRepro ledger and return its re-run outcome with its qualifiers (such as a broad bound or a check that missed), how many agent-designated controls were excluded and how many gates the test had (when it has any), how many headline claims held, the assurance level, and links to the report and evidence. Matches arXiv IDs, DOIs, URLs, titles, short names and aliases. With no exact match it returns close matches when there are any, and says the paper is not in the ledger only when there are none.
Inputs: paper (required).
The full public record for one paper: outcome and qualifiers, summary, assurance, the attempt history (released and superseded versions with reasons), every claim checked with the published value, the re-run value, the registered tolerance, the check kind and the verdict, plus evidence links.
Inputs: paper (required).
The log's final hash, the downloadable evidence bundle, the report's SHA-256 from the release manifest, and the one command that checks the bundle with only Python. The verifier checks log and artifact integrity and that the report names the log's head hash; it does not check the report's contents against the log.
Inputs: paper (required).
For a paper not yet in the ledger: returns a ready-to-send email link to OpenRepro with the request filled in. Nothing is sent or stored by this tool; the user sends the email if they want to.
Inputs: paper (required), why.
Search published re-runs by words in the title, authors, venue or field, optionally filtered by outcome or field. With no query, lists the whole ledger.
Inputs: field, limit, query, outcome.
Optional: send short feedback about the OpenRepro ledger and this server to OpenRepro's maintainers, who read it to improve the ledger. It is stored privately with the day and these fields only, kept for 12 months and never published. It must not contain personal data, the user's own words or prompts, file contents or secrets. Links, email addresses, IP addresses, phone-like numbers and long tokens are removed on a best-effort basis only. Send it only if you have feedback about the ledger, or the user is happy for it to be sent. No other tool depends on it.
Inputs: task_type (required), client, useful, suggestion, +1 more.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Check that a reference resolves; see what supports, refutes or cites a paper. CC0 graph, read-only.
Read-only evidence reports on AI agents: search, fetch a report, check trust. Not a certification.
Signed, independent scores and claim checks for decentralised AI projects. Read-only, no auth.
Search arXiv papers and read Scout reports with methods, evidence and verification points.
Add one secure layer between your agents and this server.
