Key Takeaways
- Install the B2B lead enrichment MCP server via npx to give Claude Code firmographics, technographics, and intent signals.
- Pay only when confidenceScore > 0.6, enabling cost-effective AI SDR swarms.
What Changed — A Pay-Per-Success MCP for B2B Data
A new MCP server, lead-enrichment-mcp, now gives Claude Code native access to live B2B firmographic and intent data. Unlike typical REST APIs that require custom middleware and parsing, this server exposes an enrich_lead tool directly to your agent.
The killer feature is Risk-Free Metered Billing: your account is only debited when the server returns a confidenceScore greater than 0.6. Failed lookups or low-confidence results cost exactly $0. This flips the economics of AI SDRs — you can run aggressive "wide-net" discovery across thousands of leads without burning budget on garbage data.
What It Means For You — Cost-Effective Autonomous SDRs
For Claude Code users building sales workflows, this changes two things:
No more parsing hell. The server returns strict JSON-RPC responses with Zod-schema validation, eliminating the LLM parameter hallucinations common with loosely-typed REST endpoints. Claude Code can parse
enrich_leadresults with 100% accuracy.Confidence-gated automation. Every response includes a
confidenceScore(0.0 to 1.0). You can programmatically instruct Claude Code to only proceed with outreach sequences when the score exceeds a threshold. This prevents your agent from acting on hallucinated or low-quality lead data.
The data payload includes:
- Firmographics: company size, revenue brackets, HQ location, industry
- Technographics: CRM usage, hosting providers, frontend frameworks (e.g., Salesforce, AWS, React, Marketo)
- Intent signals: hiring activity, recent funding, tech expansion
Try It Now — Setup and Usage
Add this to your claude_desktop_config.json (or the equivalent MCP settings for Claude Code):
{
"mcpServers": {
"b2b-lead-enrichment": {
"command": "npx",
"args": [
"-y",
"@modelcontextprotocol/server-http",
"--url",
"https://lead-enrichment-mcp.agent-infra.workers.dev/mcp"
],
"env": {
"API_KEY": "YOUR_ENRICHMENT_API_KEY"
}
}
}
}
Grab a free API key at lead-enrichment-mcp.agent-infra.workers.dev.
Then, in Claude Code, you can prompt:
Enrich the lead for Acme Corp. If confidenceScore > 0.7, draft a personalized outreach email mentioning their recent Series C funding and Salesforce usage.
Claude Code will call enrich_lead, parse the structured response, and act conditionally based on the confidence score.
Why This Matters — The Shift to Metered MCPs
We're seeing a broader trend in the MCP ecosystem toward stateless, metered servers that minimize context bloat and cost. This server fits that pattern: it's a remote HTTP MCP that doesn't load thousands of tool schemas into your context window. It only returns data when you explicitly call enrich_lead, keeping your agent's context lean.
This is a practical example of MCP minimalism — fewer, purpose-built servers that do one thing well, rather than monolithic integrations that slow down your agent.
Source: dev.to
