mcpField
V
verity

verity

Local
github.com/Clawdio777Search0Updated 1w ago

About verity

VERITY is a real-time fact-checking agent that checks claims, URLs, and content against live web sources. It returns structured verdicts with confidence, changes, and supporting sources, and supports single, deep, and batch checks.

Key features

  1. 01Verify claims or URLs
  2. 02Multi-angle deep checks
  3. 03Batch checks for up to 10 claims
  4. 04Structured verdicts with confidence
  5. 05Changes and supporting sources

Use cases

  1. 01Check whether facts on a webpage remain current
  2. 02Determine if content is outdated or disputed
  3. 03Verify multiple claims in one batch
  4. 04Review verdicts and their source evidence

MCP support

It exposes MCP tools for verification, deep checks, batch checks, and natural-language fact-checking, as well as a credit-balance tool.

Configuration

Local server config is the same for every client. Copy it into your client’s MCP config file.

mcp.json · Universal config
{
"mcpServers": {
"clawdio777-verity": {
"command": "npx",
"args": [
"-y",
"verity-mcp"
],
"env": {
"VERITY_PRIVATE_KEY": "<VERITY_PRIVATE_KEY>"
}
}
}
}
Run commandnpx -y verity-mcp
Environment variables to set
VERITY_PRIVATE_KEYSecretBase wallet private key (0x...) — used to sign x402 payments in USDC on Base. Per-call: 0.10 USDC (verify), 0.50 USDC (deep-check), 0.75 USDC (batch).

Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.

Packages
npmverity-mcp1.0.2

Tools

No tool list yet

Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.

Details

From the project README

VERITY — Real-Time Fact-Checking Agent

Real-time fact-checking and data freshness agent. Verifies claims, URLs, and content against live web sources. Returns structured verdicts with confidence scores, what has changed, and supporting sources.

Verdicts

VerdictMeaning
CURRENTClaim is accurate and up to date
OUTDATEDClaim was true but has since been superseded
DISPUTEDSources disagree — no clear consensus
UNVERIFIABLENo usable sources found

MCP Setup (Claude Desktop / Cursor)

{
  "mcpServers": {
    "verity": {
      "command": "npx",
      "args": ["verity-mcp"],
      "env": {
        "VERITY_PRIVATE_KEY": "0x..."
      }
    }
  }
}

Get USDC on Base at coinbase.com/wallet.

Tools

ToolDescriptionPrice
verity_verifyVerify a claim or URL against live sources0.10 USDC
verity_deep_checkMulti-angle thorough verification0.50 USDC
verity_batchVerify up to 10 claims at once0.75 USDC
verity_agentNatural language fact-checking0.10 USDC
check_creditsRemaining balance for a PayGated pg_ keyfree

API

POST https://verity.basechainlabs.com/api/verify
Authorization: x402 (USDC on Base)
{
  "claim": "Is GPT-4 still the most capable OpenAI model?",
  "caller_id": "my-agent-id"
}

Response:

{
  "verdict": "OUTDATED",
  "confidence": 91,
  "summary": "GPT-4 has been superseded by GPT-4o and o3.",
  "what_changed": "OpenAI released GPT-4o (May 2024) and o3 (late 2024).",
  "sources": [{ "url": "...", "title": "...", "published_date": "2025-01", "supports": "CONTRADICTS" }],
  "checked_at": "2026-05-11T09:00:00Z",
  "recommendation": "Update any content referencing GPT-4 as the most capable model."
}

Persistent Memory

Pass a consistent caller_id on every call. VERITY remembers topics you've checked, domains you monitor, and previous results — no re-sending context.

A2A Agent Card

GET https://verity.basechainlabs.com/api/agent?agent-card=true

Built by BaseChain Labs


Making changes (maintainer guide)

Where things live
ThingWhere
The agent's brainsrc/agent.ts (system prompt, orchestration), src/tools.ts (fact-checking: Tavily search + Claude verdicts)
API endpointsone file per route under api/: agent (paid A2A entry), verify, deep-check, batch-verify, mcp, checkout, stripe-webhook, recover-key, plus the cron-check-deps and daily-summary crons
Payment gateapi/_x402-gate.ts: unpaid POSTs get HTTP 402 with the price and the Bazaar listing; GET returns the agent card
Public pagepublic/index.html (plain HTML). Agent card: public/.well-known/agent.json
Marketplace sellerseller-v2.mjs + Dockerfile + railway.json, the Virtuals ACP seller on Railway. Its keyring lives on a Railway volume; keyring.json and keyring.key are runtime files and gitignored
Secretshosting environment variables only. Never in a file, never in a commit
How a change goes live

There is no test suite and no CI on this repo. A push to main is the deploy.

  1. Edit, then type-check: npx tsc --noEmit -p .
  2. Commit and push to main. Vercel builds and deploys production from the GitHub integration.
  3. Confirm the newest production deployment carries your commit SHA and the endpoint answers: curl -s https://verity.basechainlabs.com/.well-known/agent.json | head -c 300
  4. A push also redeploys the Railway seller. Confirm its log shows it connected afterwards.

Because there are no tests, prove a change on the live endpoint with a read-only call before calling it done. Never make a paid call (x402 or Stripe) just to test.

Things that bite
  • Prices live in four places that must stay in sync: public/index.html, the 402 body in api/_x402-gate.ts, the Virtuals ACP offerings, and the npm package README.
  • Agentic.market's "Validate endpoint" tool: use POST. GET returns the agent card (200) by design and the validator then says "no x402 setup".
  • Tavily is the search backend. If verifications start failing, check the Tavily credit first.
  • Removing the Railway volume wipes the seller's keyring on the next redeploy and takes it offline.

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.