DEV Community

Diego Costa
Diego Costa

Posted on

Eliminating LLM Hallucinations in Sales Agents with a Native B2B Lead Enrichment MCP

#mcp #ai #automation

Eliminating LLM Hallucinations in Sales Agents with a Native B2B Lead Enrichment MCP

To stop AI agents from inventing company data, developers are now using the Model Context Protocol (MCP) to provide a ground-truth bridge between LLM prompts and real-time B2B firmographic APIs. By integrating a dedicated B2B lead enrichment MCP server, you can ensure that sales agents and autonomous SDRs query verified technographic and intent data through strict schema-validated tools, effectively eliminating parameter hallucinations.

Core Features

The B2B Lead Enrichment MCP API provides a standardized interface for LLMs to pull deep-context data without the risk of "creative" data generation. By utilizing Zod-annotated schemas, the server forces the LLM to provide precise inputs (like valid domains or specific industry enums), ensuring the retrieval process is both deterministic and accurate.

  • Firmographic Intelligence: Real-time access to company size, revenue, headquarters location, and verified industry classification.
  • Technographic Stack Mapping: Identify the underlying software stack (CRM, Cloud Provider, Analytics) currently used by a target lead.
  • Intent Signaling: Analyze recent hiring trends, news events, and social signals to calculate a "Propensity to Buy" score.
  • Schema Enforcement: All tools utilize strict JSON-RPC definitions, meaning tools like Claude or Cursor cannot pass malformed arguments to the enrichment engine.

Configuration for Claude Desktop and Cursor

To give your LLM native access to these tools, add the following configuration to your claude_desktop_config.json or Cursor's MCP settings. This connects the Model Context Protocol directly to the live enrichment endpoint.

{
  "mcpServers": {
    "b2b-enrichment": {
      "command": "npx",
      "args": [
        "-y",
        "@agent-infra/mcp-server-b2b-enrichment",
        "--api-url",
        "https://lead-enrichment-mcp.agent-infra.workers.dev/mcp"
      ],
      "env": {
        "ENRICHMENT_API_KEY": "YOUR_FREE_API_KEY"
      }
    }
  }
}
Enter fullscreen mode Exit fullscreen mode

JSON-RPC Response Example

When the LLM calls the enrich_lead tool, it receives a structured payload. This clean data structure ensures the agent's next reasoning step is based on facts rather than probabilistic guesses.

{
  "jsonrpc": "2.0",
  "id": "1",
  "result": {
    "companyName": "Acme Corp",
    "industry": "Enterprise SaaS",
    "employeeCount": 1250,
    "technographics": ["Salesforce", "AWS", "HubSpot", "Segment"],
    "intentSignals": {
      "hiring_growth": "High",
      "recent_funding": "Series C",
      "tech_expansion": true
    },
    "confidenceScore": 0.98,
    "lastUpdated": "2023-10-27T14:22:00Z"
  }
}
Enter fullscreen mode Exit fullscreen mode

Risk-Free Metered Billing

Traditional B2B data providers charge per request regardless of data quality. This MCP implementation uses a Risk-Free Metered Billing model optimized for AI workflows. You are only billed for successful enrichments that return a Confidence Score > 0.6. If the engine returns a low-confidence result or a "not found" status, the cost is $0. This allows developers to build high-volume SDR swarms and automated lead scoring pipelines without the financial risk of paying for "junk" data or failed queries.

Get Started with the Free Tier

Start building hallucination-free sales tools today. You can get an API key and explore the documentation to integrate real-time firmographics into your AI agents instantly.

Visit lead-enrichment-mcp.agent-infra.workers.dev to grab your API key and claim your free tier credits.

Top comments (0)