Document Integrity Validator

Document Integrity Validator MCP

Before your agent accepts, processes, or acts on any document received from an external party -- verify it is internally consistent and matches the known standard for its type. One call. Any document. Machine-readable verdict.

What it does

Checks any document for internal consistency, completeness, and anomalies against the known international standard for that document type. Accepts base64 image or extracted text. Returns a structured verdict with a machine-readable agent_action field.

Supported standards include: ICAO Document 9303 (passports), Hague-Visby Rules 1968 (bills of lading), UCP 600 (trade finance documents), ISPM 12/IPPC/FAO (phytosanitary certificates), Vienna Convention on Road Traffic 1968 (driving licences), and more.

Returns UNKNOWN_DOCUMENT_TYPE rather than guessing on unfamiliar documents -- refusal is correct behaviour, not a failure.

AI-powered reasoning -- NOT a database lookup.

Tools

check_document (Free tier: 10 calls/month)

Checks a single document against its international standard.

Input:

• document_text (string, optional) -- extracted text from the document
• document_image (string, optional) -- base64 encoded image (raw base64 or data URL)
• document_type_hint (string, optional) -- agent belief about document type
• issuing_jurisdiction (string, optional) -- country or issuing body

At least one of document_text or document_image is required.

Response:

{

"agent_action": "PROCEED", "verdict": "PASS", "confidence": "HIGH", "document_type_identified": "Bill of Lading", "assessed_against": "Hague-Visby Rules 1968", "known_issuing_standard": "IMO", "flags": [], "reason": "Document is internally consistent and compliant with Hague-Visby Rules 1968.", "analysis_type": "AI-powered reasoning -- NOT a database lookup", "checked_at": "2026-05-06T10:00:00.000Z", "_disclaimer": "..." }

agent_action values:

• PROCEED -- document passed
• VERIFY_MANUALLY -- flags found, agent should flag for human review
• HOLD -- document failed, do not proceed
• REFER_TO_HUMAN -- document type unknown, refer for manual assessment

check_document_package (Paid tier only)

Checks 2-20 related documents individually then cross-checks all for consistency conflicts.

Input:

• documents (array, min 2, max 20) -- each item has: label (required), document_text, document_image, document_type_hint, issuing_jurisdiction

Cross-checks performed: weights/quantities/amounts, party names, reference numbers, dates, commodity descriptions, port references.

Pricing

Overage: $0.05 per call above monthly cap.

Subscribe at kordagencies.com.

Harness Integration

Claude Code / Claude Desktop (.mcp.json)

{

"mcpServers": { "document-integrity-validator": { "type": "http", "url": "https://document-integrity-validator-mcp-production.up.railway.app/mcp" } } }

LangChain / LangGraph (Python)

from langchain_mcp_adapters.client import MultiServerMCPClient

client = MultiServerMCPClient({ "document-integrity-validator": { "url": "https://document-integrity-validator-mcp-production.up.railway.app/mcp", "transport": "http" } }) tools = await client.get_tools()

OpenAI Agents SDK (Python)

from agents import Agent, HostedMCPTool

agent = Agent( name="Assistant", tools=[HostedMCPTool(tool_config={ "type": "mcp", "server_label": "document-integrity-validator", "server_url": "https://document-integrity-validator-mcp-production.up.railway.app/mcp", "require_approval": "never" })] )

Self-hosted (stdio)

npm install -g document-integrity-validator-mcp

ANTHROPIC_API_KEY=sk-ant-... TRANSPORT=stdio document-integrity-validator-mcp

Add to Claude Desktop claude_desktop_config.json:

{

"mcpServers": { "document-integrity-validator": { "command": "document-integrity-validator-mcp", "env": { "ANTHROPIC_API_KEY": "sk-ant-..." } } } }

Trial Extension

If you have reached the 10 call/month free limit, request 10 extra calls:

curl -X POST https://document-integrity-validator-mcp-production.up.railway.app/trial-extension \

-H "Content-Type: application/json" \ -d '{"name":"Your Name","email":"you@example.com","use_case":"Brief description"}'

One extension per email address.

Legal

AI-powered document consistency assessment. Results are for informational purposes only and do not constitute legal, compliance, or authentication advice. We do not log or store your document content. Full terms: kordagencies.com/terms.html