Skip to main content
Glama
anish632

ground-truth-mcp

by anish632
OverviewSchemaRelated ServersScoreDiscussions
TypeScript
Remote

Claim Support Check

verify_claim
Read-onlyIdempotent

Verify a factual claim against provided public evidence URLs using keyword matching. Input claim, URLs, and keywords; each source is marked supporting when at least half the keywords appear. Use for evidence-backed checks on known pages.

Instructions

Check whether a factual claim is supported by a specific set of public evidence URLs that you already have. For each source, the tool performs a case-insensitive keyword match over the fetched page body, then marks that source as supporting the claim when at least half of the supplied keywords appear. Use this for evidence-backed claim checks on known pages, not for open-ended search or semantic fact checking. Registry responses are cached for 5 minutes.

Input Schema

TableJSON Schema
NameRequiredDescriptionDefault
claimYesPlain-language claim to verify, for example 'AWS Business support includes 24/7 phone support'.
evidence_urlsYesOne to ten public documentation, pricing, policy, or support URLs that are likely to contain direct evidence for the claim.
keywordsYesKeywords or short phrases that should appear on supporting pages. Matching is case-insensitive substring matching.

Output Schema

TableJSON Schema
NameRequiredDescriptionDefault
claimYesClaim that was evaluated.
sourcesYesPer-source evidence results.
verdictYesAggregate verdict across all supplied sources.

Implementation Reference

  • src/index.ts:975-1094 (registration)
    Registration of the 'verify_claim' tool via this.server.registerTool(), including its input/output schema definitions and the handler function.
    this.server.registerTool(
  "verify_claim",
  {
    title: "Claim Support Check",
    description:
      "Check whether a factual claim is supported by a specific set of public " +
      "evidence URLs that you already have. For each source, the tool performs a " +
      "case-insensitive keyword match over the fetched page body, then marks that " +
      "source as supporting the claim when at least half of the supplied keywords " +
      "appear. Use this for evidence-backed claim checks on known pages, not for " +
      "open-ended search, semantic reasoning, or contradiction extraction. The " +
      "aggregate verdict is driven only by the per-page keyword support ratio. " +
      "Fetched pages are cached for 5 minutes.",
    inputSchema: {
      claim: z.string().trim().min(5).describe(
        "Plain-language claim to verify, for example 'AWS Business support includes 24/7 phone support'.",
      ),
      evidence_urls: z.array(z.string().url()).min(1).max(10).describe(
        "One to ten public documentation, pricing, policy, or support URLs that are likely to contain direct evidence for the claim.",
      ),
      keywords: z.array(z.string().trim().min(1)).min(1).max(20).describe(
        "Keywords or short phrases that should appear on supporting pages. Matching is case-insensitive substring matching, so choose phrases that are likely to appear verbatim.",
      ),
    },
    outputSchema: {
      claim: z.string().describe(
        "Claim that was evaluated.",
      ),
      sources: z.array(z.object({
        url: z.string().describe(
          "Evidence URL that was checked.",
        ),
        accessible: z.boolean().describe(
          "True when the evidence page could be fetched.",
        ),
        cached: z.boolean().describe(
          "True when the page body came from the 5-minute cache.",
        ).optional(),
        keywordsMatched: z.array(z.string()).describe(
          "Subset of supplied keywords that were found on the page.",
        ).optional(),
        keywordsTotal: z.number().int().nonnegative().describe(
          "Total number of keywords the tool looked for on this page.",
        ).optional(),
        matchRatio: z.number().min(0).max(1).describe(
          "Matched-keyword ratio for this source, from 0 to 1.",
        ).optional(),
        supports: z.boolean().describe(
          "True when the page met the current support threshold of at least half of the supplied keywords.",
        ),
        error: z.string().describe(
          "Fetch error when the evidence page could not be checked.",
        ).optional(),
      })).describe(
        "Per-source evidence results.",
      ),
      verdict: z.object({
        supporting: z.number().int().nonnegative().describe(
          "Number of sources marked as supporting the claim.",
        ),
        contradicting: z.number().int().nonnegative().describe(
          "Number of sources not marked as supporting the claim.",
        ),
        total: z.number().int().nonnegative().describe(
          "Total number of evidence sources checked.",
        ),
        confidence: z.number().min(0).max(1).describe(
          "Share of sources that supported the claim.",
        ),
        summary: z.enum(["CONFIRMED", "UNCONFIRMED", "LIKELY TRUE", "LIKELY FALSE"]).describe(
          "High-level verdict derived from the supporting-source ratio: all sources supporting => CONFIRMED, none => UNCONFIRMED, majority => LIKELY TRUE, otherwise LIKELY FALSE.",
        ),
      }).describe(
        "Aggregate verdict across all supplied sources.",
      ),
    },
    annotations: readOnlyNetworkToolAnnotations,
  },
  async ({ claim, evidence_urls, keywords }) => {
    const sources = [];
    for (const url of evidence_urls) {
      try {
        const { body, fromCache } = await cachedFetch(sql, url);
        const bodyLower = body.toLowerCase();
        const keywordHits = keywords.filter(kw => bodyLower.includes(kw.toLowerCase()));
        sources.push({
          url,
          accessible: true,
          cached: fromCache,
          keywordsMatched: keywordHits,
          keywordsTotal: keywords.length,
          matchRatio: +(keywordHits.length / keywords.length).toFixed(2),
          supports: keywordHits.length >= keywords.length * 0.5,
        });
      } catch (e: unknown) {
        sources.push({
          url,
          accessible: false,
          error: e instanceof Error ? e.message : String(e),
          supports: false,
        });
      }
    }
    const supporting = sources.filter(s => s.supports).length;
    logUsage("verify_claim", true);
    return structuredToolResult({
      claim,
      sources,
      verdict: {
        supporting,
        contradicting: sources.length - supporting,
        total: sources.length,
        confidence: +(supporting / sources.length).toFixed(2),
        summary: supporting === sources.length ? "CONFIRMED" :
                 supporting === 0 ? "UNCONFIRMED" :
                 supporting >= sources.length * 0.5 ? "LIKELY TRUE" : "LIKELY FALSE",
      },
    });
  }
);
  • src/index.ts:1053-1094 (handler)
    Handler function for 'verify_claim' that fetches each evidence URL, performs case-insensitive substring matching of keywords against the page body, marks a source as supporting if at least half of keywords match, and produces an aggregate verdict (CONFIRMED, UNCONFIRMED, LIKELY TRUE, LIKELY FALSE).
      async ({ claim, evidence_urls, keywords }) => {
    const sources = [];
    for (const url of evidence_urls) {
      try {
        const { body, fromCache } = await cachedFetch(sql, url);
        const bodyLower = body.toLowerCase();
        const keywordHits = keywords.filter(kw => bodyLower.includes(kw.toLowerCase()));
        sources.push({
          url,
          accessible: true,
          cached: fromCache,
          keywordsMatched: keywordHits,
          keywordsTotal: keywords.length,
          matchRatio: +(keywordHits.length / keywords.length).toFixed(2),
          supports: keywordHits.length >= keywords.length * 0.5,
        });
      } catch (e: unknown) {
        sources.push({
          url,
          accessible: false,
          error: e instanceof Error ? e.message : String(e),
          supports: false,
        });
      }
    }
    const supporting = sources.filter(s => s.supports).length;
    logUsage("verify_claim", true);
    return structuredToolResult({
      claim,
      sources,
      verdict: {
        supporting,
        contradicting: sources.length - supporting,
        total: sources.length,
        confidence: +(supporting / sources.length).toFixed(2),
        summary: supporting === sources.length ? "CONFIRMED" :
                 supporting === 0 ? "UNCONFIRMED" :
                 supporting >= sources.length * 0.5 ? "LIKELY TRUE" : "LIKELY FALSE",
      },
    });
  }
);
  • src/index.ts:977-1051 (schema)
    Input/output schema definitions for 'verify_claim': takes a claim, evidence_urls (1-10 URLs), and keywords (1-20 phrases); returns per-source match results and an aggregate verdict.
    {
  title: "Claim Support Check",
  description:
    "Check whether a factual claim is supported by a specific set of public " +
    "evidence URLs that you already have. For each source, the tool performs a " +
    "case-insensitive keyword match over the fetched page body, then marks that " +
    "source as supporting the claim when at least half of the supplied keywords " +
    "appear. Use this for evidence-backed claim checks on known pages, not for " +
    "open-ended search, semantic reasoning, or contradiction extraction. The " +
    "aggregate verdict is driven only by the per-page keyword support ratio. " +
    "Fetched pages are cached for 5 minutes.",
  inputSchema: {
    claim: z.string().trim().min(5).describe(
      "Plain-language claim to verify, for example 'AWS Business support includes 24/7 phone support'.",
    ),
    evidence_urls: z.array(z.string().url()).min(1).max(10).describe(
      "One to ten public documentation, pricing, policy, or support URLs that are likely to contain direct evidence for the claim.",
    ),
    keywords: z.array(z.string().trim().min(1)).min(1).max(20).describe(
      "Keywords or short phrases that should appear on supporting pages. Matching is case-insensitive substring matching, so choose phrases that are likely to appear verbatim.",
    ),
  },
  outputSchema: {
    claim: z.string().describe(
      "Claim that was evaluated.",
    ),
    sources: z.array(z.object({
      url: z.string().describe(
        "Evidence URL that was checked.",
      ),
      accessible: z.boolean().describe(
        "True when the evidence page could be fetched.",
      ),
      cached: z.boolean().describe(
        "True when the page body came from the 5-minute cache.",
      ).optional(),
      keywordsMatched: z.array(z.string()).describe(
        "Subset of supplied keywords that were found on the page.",
      ).optional(),
      keywordsTotal: z.number().int().nonnegative().describe(
        "Total number of keywords the tool looked for on this page.",
      ).optional(),
      matchRatio: z.number().min(0).max(1).describe(
        "Matched-keyword ratio for this source, from 0 to 1.",
      ).optional(),
      supports: z.boolean().describe(
        "True when the page met the current support threshold of at least half of the supplied keywords.",
      ),
      error: z.string().describe(
        "Fetch error when the evidence page could not be checked.",
      ).optional(),
    })).describe(
      "Per-source evidence results.",
    ),
    verdict: z.object({
      supporting: z.number().int().nonnegative().describe(
        "Number of sources marked as supporting the claim.",
      ),
      contradicting: z.number().int().nonnegative().describe(
        "Number of sources not marked as supporting the claim.",
      ),
      total: z.number().int().nonnegative().describe(
        "Total number of evidence sources checked.",
      ),
      confidence: z.number().min(0).max(1).describe(
        "Share of sources that supported the claim.",
      ),
      summary: z.enum(["CONFIRMED", "UNCONFIRMED", "LIKELY TRUE", "LIKELY FALSE"]).describe(
        "High-level verdict derived from the supporting-source ratio: all sources supporting => CONFIRMED, none => UNCONFIRMED, majority => LIKELY TRUE, otherwise LIKELY FALSE.",
      ),
    }).describe(
      "Aggregate verdict across all supplied sources.",
    ),
  },
  annotations: readOnlyNetworkToolAnnotations,

Tool Definition Quality

A5/5.0
Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Description adds detail beyond annotations: case-insensitive substring keyword match, half-keyword threshold for support, 5-minute cache. Does not contradict annotations.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Two sentences, front-loaded with purpose, no redundant phrases. Every sentence adds value.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Fully covers purpose, usage, behavior, parameters, and caching. Provides enough context for an AI agent to select and invoke correctly, especially given the rich annotations and output schema.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters5/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema has 100% coverage, but description adds crucial details: case-insensitive matching, threshold of half keywords, and context that evidence_urls should be public documentation/pricing/policy/support URLs.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description clearly states the tool checks whether a claim is supported by specific URLs using keyword matching. It distinguishes from siblings by specifying it is for evidence-backed checks on known pages, not open-ended search.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Explicitly states when to use (evidence-backed checks on known pages) and when not to use (open-ended search or semantic fact checking). Also mentions caching behavior.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Install Server

Other Tools

Latest Blog Posts

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/anish632/ground-truth-mcp'

If you have feedback or need assistance with the MCP directory API, please join our Discord server