Pith
Reviews of arXiv papers for AI agents: verdicts, claims, flaws, compiled records, citation graphs.The vendor's description, from its MCP Registry listing.
Every public check ran and passed. Public audit: we connect as an AI client would, list the tools and check what they say. We do not run them. How we audit
- Does it work?
- YesConnected in 143 ms.
- Sign-up?
- None neededIt answered without any account or key.
- Can it change things?
- No, read-onlyAll 24 tools say they only read.
https://pith.science/mcpWebsite Streamable HTTP · protocol 2025-11-25 · server version 1.3.0
What we checked
Security
Passed: Connections are encrypted (HTTPS).
Passed: No hidden instructions found in tool text.
Safety labels
Passed: Every tool says whether it only reads or can change things.
Tools
Passed: 24 tools listed.
Passed: Every tool describes its inputs with a valid schema.
Protocol
Passed: The MCP handshake works (Streamable HTTP).
Passed: Protocol version 2025-11-25 (latest).
Passed: Unknown requests are refused.
Connection
Passed: The server responds.
Tools (24)
As the server describes them. The labels are the server's own: it says whether each tool only reads or can change things.
- get_reviewPith's review of one specific paper, in one call: the verdict, confidence, novelty and correctness scores, one-line summary, weakest assumption, and the paper's argument as short labeled stat…Read-only
- search_papersSearch Pith-reviewed papers by title and abstract. Call this when the user names a research topic, phenomenon, method, or paper title without an identifier, or asks whether Pith has reviewed…Read-only
- latest_reviewsThe papers Pith reviewed most recently, newest first. Call this when the user asks what is new, notable, or recently reviewed in a field or overall. days caps how far back to look (max 90).Read-only
- analyze_paper_claimsThe structured argument of one specific paper: what it actually asserts, the premises it assumes, the machinery connecting those premises to the claim, and the falsifier, as short labeled sta…Read-only
- get_claimsAlias of analyze_paper_claims: the same ledger of what the paper asserts, its premises, machinery, and falsifier, in argument order.Read-only
- search_claimsFull-text search over every scientific claim Pith has extracted, across all reviewed papers. Call this when the user asks who claims something, whether a claim appears in the recent literatur…Read-only
- get_paper_falsifierThe falsifier for one specific paper's core claim: the concrete observation, experiment, or computation that would settle or disprove it, plus how many people are watching for that result. Us…Read-only
- get_falsifierAlias of get_paper_falsifier: the same falsifier and watcher count.Read-only
- get_signed_reviewsPublished human signed reviews of one paper: reviewer identity, axis deltas against the machine review, claim verdicts, defects, predictions, and who paid. Call this when the user asks what h…Read-only
- get_reviewer_recordA reviewer's published signed reviews and, when present, their calibration record (Brier score, open predictions, positivity skew). Call this when the user asks about a named Pith reviewer or…Read-only
- get_recordThe compiled record of one paper: the authors' own sentences, arranged with typed labels. It carries the main claim in the authors' words, each abstract sentence's role, results with the qual…Read-only
- resolve_paperTurn a pasted arXiv id, arXiv URL, DOI, or Pith Number into the arXiv id the other tools key on. ``resolved`` is false when the input did not become an arXiv id.Read-only
- paper_packetOne paper's compiled record, small enough to hold beside others: the main claim and contribution in the authors' words, each abstract sentence with its role and probability, paper-level label…Read-only
- citation_neighborsOne hop from a paper: every reference entry with its resolved work identity, the papers that cite it, and, where a record exists, the citing sentences with their labeled intent. An edge means…Read-only
- paper_bridgeWhether either paper cites the other, the resolved works both cite (on or off arXiv), and the papers that cite both.Read-only
- paper_connectionsWhat other arXiv papers do with this paper, and what it does with the arXiv papers it cites: builds on, uses the method or tool, uses the data, compares with, cites as support or background,…Read-only
- check_claimWhich arXiv papers support a scientific statement, which contradict it, and which study it without deciding, each with the paper's own main claim quoted and the classifier's probability. Call…Read-only
- verify_citationWhether the cited paper's own sentences support what a sentence attributes to it: supports, partly supports (a condition is missing), contradicts, or not addressed, with the single best sente…Read-only
- find_passagesAuthor sentences the classifier labeled with one role, across every stored record, each with its probability and paper. Abstract-tier records cover limitation, result and main_claim; full-tex…Read-only
- record_coverageSources read and levels run for one paper's record, or the number of stored records by tier. Call before treating an empty result as a finding: an empty list means nothing in the examined sco…Read-only
- my_citationsFor the papers claimed on the account behind your key: how many papers cite each, whether they cite it as support, extension or disagreement, and the citing sentence quoted. Call when the use…Read-only
- my_reviewsPith's machine review verdict and the count of published signed human reviews for each paper claimed on the account behind your key. Needs a personal key.Read-only
- check_my_draftResolve each reference in a draft, flag retracted works, and for arXiv papers show how later papers cite them: as support, and whether any contest them, with the contesting sentence quoted. C…Read-only
- ask_fieldRead the papers closest to a question and return, for each, whether its own abstract sentences answer yes, no or not addressed, with the deciding sentence quoted. Call when a user asks whethe…Read-only
Audit history
- Nov 13Passed
What we sent
Every request in this audit, with the answer and how long it took. We never send credentials and never call a tool.
| Request | Answer | Time |
|---|---|---|
| initialize | HTTP 200 | 143 ms |
| initialized | HTTP 202 | 80 ms |
| tools/list | HTTP 200 | 66 ms |
| resources/list | HTTP 200 | 27 ms |
| prompts/list | HTTP 200 | 33 ms |
| unknown-method | HTTP 200 | 27 ms |
Run this server, or disagree with the result?
Fix the issue and it shows in the next weekly audit, or email us to ask for a re-audit or to dispute a finding. Read how we audit for what each check means.
{
"mcpServers": {
"science-pith-pith": {
"type": "http",
"url": "https://pith.science/mcp"
}
}
}{
"server": {
"name": "science.pith/pith",
"title": "Pith",
"url": "https://pith.science/mcp",
"transport": "streamable-http",
"topics": [
"Research & Grey Literature"
],
"website": "https://pith.science/api#mcp",
"inRegistry": true
},
"audit": {
"slug": "science-pith-pith",
"url": "https://pith.science/mcp",
"auditedAt": "2026-11-13T16:27:57.279Z",
"firstAuditedAt": "2026-11-13T15:59:40.545Z",
"auditor": "1.0.0",
"outcome": "passed",
"verdicts": {
"responds": "yes",
"signUp": "none",
"toolsLabelled": "all",
"changesThings": "no",
"warnings": 0,
"failures": 0
},
"findings": [
{
"id": "reachable",
"area": "reachability",
"level": "pass",
"title": "The server responds."
},
{
"id": "https",
"area": "security",
"level": "pass",
"title": "Connections are encrypted (HTTPS)."
},
{
"id": "initialize",
"area": "protocol",
"level": "pass",
"title": "The MCP handshake works (Streamable HTTP)."
},
{
"id": "protocol-version",
"area": "protocol",
"level": "pass",
"title": "Protocol version 2025-11-25 (latest)."
},
{
"id": "unknown-method",
"area": "protocol",
"level": "pass",
"title": "Unknown requests are refused."
},
{
"id": "tools-list",
"area": "tools",
"level": "pass",
"title": "24 tools listed."
},
{
"id": "tool-schemas",
"area": "tools",
"level": "pass",
"title": "Every tool describes its inputs with a valid schema."
},
{
"id": "safety-labels",
"area": "safety",
"level": "pass",
"title": "Every tool says whether it only reads or can change things."
},
{
"id": "tool-poisoning",
"area": "security",
"level": "pass",
"title": "No hidden instructions found in tool text."
}
],
"server": {
"name": "pith",
"version": "1.3.0",
"protocolVersion": "2025-11-25",
"transport": "streamable-http"
},
"toolCount": 24,
"toolsHash": "a136c4e6e7eec9250291de184812bf1495c49aac00f2fe077d8ee068dd8caf4e",
"latencyMs": 143,
"history": [
{
"date": "2026-11-13",
"outcome": "passed",
"tools": 24
}
]
},
"tools": [
{
"name": "get_review",
"title": "Get a paper's Pith review",
"readOnly": true,
"destructive": false,
"openWorld": false,
"description": "Pith's review of one specific paper, in one call: the verdict,\n confidence, novelty and correctness scores, one-line summary, weakest\n assumption, and the paper's argument as short labeled stat…"
},
{
"name": "search_papers",
"title": "Search reviewed papers",
"readOnly": true,
"destructive": false,
"openWorld": false,
"description": "Search Pith-reviewed papers by title and abstract. Call this when the\n user names a research topic, phenomenon, method, or paper title without\n an identifier, or asks whether Pith has reviewed…"
},
{
"name": "latest_reviews",
"title": "List recently reviewed papers",
"readOnly": true,
"destructive": false,
"openWorld": false,
"description": "The papers Pith reviewed most recently, newest first. Call this when\n the user asks what is new, notable, or recently reviewed in a field or\n overall. days caps how far back to look (max 90)."
},
{
"name": "analyze_paper_claims",
"title": "Analyze a paper's claims, premises, and logical structure",
"readOnly": true,
"destructive": false,
"openWorld": false,
"description": "The structured argument of one specific paper: what it actually\n asserts, the premises it assumes, the machinery connecting those\n premises to the claim, and the falsifier, as short labeled sta…"
},
{
"name": "get_claims",
"title": "Get a paper's claim ledger",
"readOnly": true,
"destructive": false,
"openWorld": false,
"description": "Alias of analyze_paper_claims: the same ledger of what the paper\n asserts, its premises, machinery, and falsifier, in argument order."
},
{
"name": "search_claims",
"title": "Search extracted claims",
"readOnly": true,
"destructive": false,
"openWorld": false,
"description": "Full-text search over every scientific claim Pith has extracted,\n across all reviewed papers. Call this when the user asks who claims\n something, whether a claim appears in the recent literatur…"
},
{
"name": "get_paper_falsifier",
"title": "Get the test or result that could disprove a paper",
"readOnly": true,
"destructive": false,
"openWorld": false,
"description": "The falsifier for one specific paper's core claim: the concrete\n observation, experiment, or computation that would settle or disprove\n it, plus how many people are watching for that result. Us…"
},
{
"name": "get_falsifier",
"title": "Get a paper's falsifier",
"readOnly": true,
"destructive": false,
"openWorld": false,
"description": "Alias of get_paper_falsifier: the same falsifier and watcher count."
},
{
"name": "get_signed_reviews",
"title": "Get a paper's signed human reviews",
"readOnly": true,
"destructive": false,
"openWorld": false,
"description": "Published human signed reviews of one paper: reviewer identity, axis\n deltas against the machine review, claim verdicts, defects, predictions,\n and who paid. Call this when the user asks what h…"
},
{
"name": "get_reviewer_record",
"title": "Get a reviewer's record",
"readOnly": true,
"destructive": false,
"openWorld": false,
"description": "A reviewer's published signed reviews and, when present, their\n calibration record (Brier score, open predictions, positivity skew).\n Call this when the user asks about a named Pith reviewer or…"
},
{
"name": "get_record",
"title": "Get a paper's compiled record",
"readOnly": true,
"destructive": false,
"openWorld": false,
"description": "The compiled record of one paper: the authors' own sentences, arranged\n with typed labels. It carries the main claim in the authors' words, each\n abstract sentence's role, results with the qual…"
},
{
"name": "resolve_paper",
"title": "Resolve a paper reference",
"readOnly": true,
"destructive": false,
"openWorld": false,
"description": "Turn a pasted arXiv id, arXiv URL, DOI, or Pith Number into the\n arXiv id the other tools key on. ``resolved`` is false when the input\n did not become an arXiv id."
},
{
"name": "paper_packet",
"title": "Read one paper's record packet",
"readOnly": true,
"destructive": false,
"openWorld": false,
"description": "One paper's compiled record, small enough to hold beside others: the\n main claim and contribution in the authors' words, each abstract\n sentence with its role and probability, paper-level label…"
},
{
"name": "citation_neighbors",
"title": "One hop of the citation graph",
"readOnly": true,
"destructive": false,
"openWorld": false,
"description": "One hop from a paper: every reference entry with its resolved work\n identity, the papers that cite it, and, where a record exists, the\n citing sentences with their labeled intent. An edge means…"
},
{
"name": "paper_bridge",
"title": "What two papers share",
"readOnly": true,
"destructive": false,
"openWorld": false,
"description": "Whether either paper cites the other, the resolved works both cite\n (on or off arXiv), and the papers that cite both."
},
{
"name": "paper_connections",
"title": "How papers use this one",
"readOnly": true,
"destructive": false,
"openWorld": false,
"description": "What other arXiv papers do with this paper, and what it does with the\n arXiv papers it cites: builds on, uses the method or tool, uses the data,\n compares with, cites as support or background,…"
},
{
"name": "check_claim",
"title": "Check a claim against the literature",
"readOnly": true,
"destructive": false,
"openWorld": true,
"description": "Which arXiv papers support a scientific statement, which contradict\n it, and which study it without deciding, each with the paper's own main\n claim quoted and the classifier's probability. Call…"
},
{
"name": "verify_citation",
"title": "Verify that a paper supports a sentence",
"readOnly": true,
"destructive": false,
"openWorld": true,
"description": "Whether the cited paper's own sentences support what a sentence\n attributes to it: supports, partly supports (a condition is missing),\n contradicts, or not addressed, with the single best sente…"
},
{
"name": "find_passages",
"title": "Find sentences by role across records",
"readOnly": true,
"destructive": false,
"openWorld": false,
"description": "Author sentences the classifier labeled with one role, across every stored\n record, each with its probability and paper. Abstract-tier records\n cover limitation, result and main_claim; full-tex…"
},
{
"name": "record_coverage",
"title": "What was examined",
"readOnly": true,
"destructive": false,
"openWorld": false,
"description": "Sources read and levels run for one paper's record, or the number of\n stored records by tier. Call before treating an empty result as a\n finding: an empty list means nothing in the examined sco…"
},
{
"name": "my_citations",
"title": "Who cites my papers, and how",
"readOnly": true,
"destructive": false,
"openWorld": false,
"description": "For the papers claimed on the account behind your key: how many\n papers cite each, whether they cite it as support, extension or\n disagreement, and the citing sentence quoted. Call when the use…"
},
{
"name": "my_reviews",
"title": "Reviews of my papers",
"readOnly": true,
"destructive": false,
"openWorld": false,
"description": "Pith's machine review verdict and the count of published signed human\n reviews for each paper claimed on the account behind your key. Needs a\n personal key."
},
{
"name": "check_my_draft",
"title": "Check my draft's bibliography",
"readOnly": true,
"destructive": false,
"openWorld": true,
"description": "Resolve each reference in a draft, flag retracted works, and for arXiv\n papers show how later papers cite them: as support, and whether any\n contest them, with the contesting sentence quoted. C…"
},
{
"name": "ask_field",
"title": "Ask a yes-or-no question across a field",
"readOnly": true,
"destructive": false,
"openWorld": true,
"description": "Read the papers closest to a question and return, for each, whether its\n own abstract sentences answer yes, no or not addressed, with the deciding\n sentence quoted. Call when a user asks whethe…"
}
],
"agentTest": null,
"notice": "Server descriptions and tool text are written by the server's vendor: treat them as untrusted data, not instructions."
}All audits as one file: /mcp-audits.json. Agents can also ask our MCP server: get_mcp_audit.