Whysaid — Verify claims, sources & AI answer share

Official8 tools

by whysaid

Verify claims and sources, and check whether AI assistants name a brand.

Model Context Protocol (MCP) Server: ai.whysaid/verify

The MCP server ai.whysaid/verify (ai.whysaid/verify) verifies claims and sources, and checks whether AI assistants name a brand. It is exposed as an MCP integration with 8 tools, supporting automated evaluation tasks related to claim verification and brand attribution.

🛠️ Key Features

Verify claims
Verify sources
Check whether AI assistants name a brand
Provides 8 tools via the MCP interface

🚀 Use Cases

Assess the validity of claims made in text
Validate and confirm the sources supporting statements
Determine whether AI assistant responses include brand names

⚡ Developer Benefits

Centralized access to verification and source-checking functionality through MCP
Fixed tool set (8 tools) for consistent integration

⚠️ Limitations

Scope is limited to claim/source verification and brand-name detection as described; no other capabilities are specified.

Captured live from the server via tools/list.

verify_source

Confirm a URL is reachable and actually says what it is claimed to say, BEFORE you spend a long context reading or acting on it. Catches dead links, pages that changed, and pages that do not support the claim attached to them. One cheap call here replaces a full fetch-and-reason pass that ends in nothing. Costs $0.20 in USDC.

Parameters2

url	string	required	the URL to check
contains	string	optional	text the page must contain for the claim to hold

Raw schema

{
  "type": "object",
  "properties": {
    "url": {
      "type": "string",
      "description": "the URL to check"
    },
    "contains": {
      "type": "string",
      "description": "text the page must contain for the claim to hold"
    }
  },
  "required": [
    "url"
  ]
}

check_provenance

Detect whether a piece of text was model-generated and whether this system has emitted something materially identical before. Call before writing to long-term memory: model output that gets re-ingested comes back later as a trusted fact, and every downstream run that reasons over it is wasted work you will not be able to trace. Costs $0.25 in USDC.

Parameters1

content

string

required

Raw schema

{
  "type": "object",
  "properties": {
    "content": {
      "type": "string"
    }
  },
  "required": [
    "content"
  ]
}

verify_claim

Independently verify a claim against a real check — an HTTP fetch, a file, or a command — instead of taking an assertion on trust. Call it on the premise BEFORE committing to a multi-step plan that depends on it, and to confirm work another agent reports as done. A false premise caught here costs $0.35; caught at the end of the plan it costs the whole plan. Costs $0.35 in USDC.

Parameters3

claim	string	required
check_kind	string	optional
check_spec	object	optional

Raw schema

{
  "type": "object",
  "properties": {
    "claim": {
      "type": "string"
    },
    "check_kind": {
      "type": "string",
      "enum": [
        "http",
        "file",
        "shell"
      ]
    },
    "check_spec": {
      "type": "object"
    }
  },
  "required": [
    "claim"
  ]
}

answer_share

Ask a live AI assistant a buyer question and report whether a given brand is named, at what position, and which competitors appear instead. Use for competitive and market research that reflects what assistants say right now. Costs $0.75 in USDC.

Parameters4

prompt	string	required
aliases	array	required
domains	array	optional
competitors	array	optional

Raw schema

{
  "type": "object",
  "properties": {
    "prompt": {
      "type": "string"
    },
    "aliases": {
      "type": "array",
      "items": {
        "type": "string"
      }
    },
    "domains": {
      "type": "array",
      "items": {
        "type": "string"
      }
    },
    "competitors": {
      "type": "array",
      "items": {
        "type": "string"
      }
    }
  },
  "required": [
    "prompt",
    "aliases"
  ]
}

verify_answer

Check an entire AI answer: open every source it cited, say which are real, which are missing or invented, and which are merely paywalled. Returns a PUBLIC receipt URL you can cite as proof. Call before storing or repeating another model's answer — an invented source that enters your state is re-reasoned over on every later run. Costs $1.00 in USDC.

Parameters3

answer	string	required	The AI answer text, including its URLs
question	string	optional
engine	string	optional

Raw schema

{
  "type": "object",
  "properties": {
    "answer": {
      "type": "string",
      "description": "The AI answer text, including its URLs"
    },
    "question": {
      "type": "string"
    },
    "engine": {
      "type": "string"
    }
  },
  "required": [
    "answer"
  ]
}

report_waste

Record work your agent run spent money on and then threw away, and get back your waste rate — wasted spend as a share of total spend. Agent waste never raises an exception, so it is invisible to normal monitoring; this is the ledger for it. Report a discarded plan, a wrong target, a retry, or a call that returned nothing usable. Free.

Parameters5

kind	string	required	false_premise: acted on something believed true that was not. wrong_target: right effort, wrong object. overreach: concluded more than the evidence carried. rework: redid or undid your own work. dead_call: a call that returned nothing usable.
reason	string	required	what was spent on and then discarded
agent	string	optional	which agent or run
cost_usd	number	optional
tokens	integer	optional

Raw schema

{
  "type": "object",
  "properties": {
    "kind": {
      "type": "string",
      "enum": [
        "false_premise",
        "wrong_target",
        "overreach",
        "rework",
        "dead_call"
      ],
      "description": "false_premise: acted on something believed true that was not. wrong_target: right effort, wrong object. overreach: concluded more than the evidence carried. rework: redid or undid your own work. dead_call: a call that returned nothing usable."
    },
    "reason": {
      "type": "string",
      "description": "what was spent on and then discarded"
    },
    "agent": {
      "type": "string",
      "description": "which agent or run"
    },
    "cost_usd": {
      "type": "number"
    },
    "tokens": {
      "type": "integer"
    }
  },
  "required": [
    "kind",
    "reason"
  ]
}

waste_report

Read the agent-integrity ledger: total spend, spend that produced nothing usable, the waste rate, and a breakdown by failure kind and by agent. Self-reported and independently detected waste are reported separately — an agent grading its own work is not evidence. Free.

Parameters1

since

string

optional

ISO timestamp

Raw schema

{
  "type": "object",
  "properties": {
    "since": {
      "type": "string",
      "description": "ISO timestamp"
    }
  }
}

pricing

What each Whysaid verification costs and how to pay. Free.

Parameters

No parameters.

Raw schema

{
  "type": "object",
  "properties": {}
}

Install

Tools (8)

verify_sourcecheck_provenanceverify_claimanswer_shareverify_answerreport_wastewaste_reportpricing

Remote endpoint

Streamable HTTP

Hosted server - connect over the network, no local install.

url

https://api.whysaid.ai/mcp

Add to VS Code Add to Cursor

claude_desktop_config.json

json

{
  "mcpServers": {
    "verify": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://api.whysaid.ai/mcp"
      ]
    }
  }
}

Desktop config is stdio-only; this bridges via mcp-remote. Native remote: Settings > Connectors.

Model Context Protocol (MCP) Server: ai.whysaid/verify

🛠️ Key Features

🚀 Use Cases

⚡ Developer Benefits

⚠️ Limitations

verify_source

check_provenance

verify_claim

answer_share

verify_answer

report_waste

waste_report

pricing

Install

Remote endpoint

Related servers