Fundnode, the AI-native referral platform for US small-business funding, today released @fundnode/mcp-server, an open Model Context Protocol (MCP) server that exposes the company's merchant cash advance underwriting database to AI agents. The package is available on npm and has been submitted to the official Model Context Protocol registry.
The release makes Fundnode the first transparent MCA platform to ship a production MCP server. Any AI agent that speaks the Model Context Protocol — including Anthropic's Claude, OpenAI's ChatGPT with MCP support, and emerging agentic IDEs — can now query Fundnode's 100-funder database, run factor-rate math, score a hypothetical merchant against real funder underwriting boxes, and surface the current state of US commercial financing disclosure law.
“If a small-business owner asks Claude or ChatGPT 'who would actually fund my restaurant with 8 months in business and $40K monthly revenue,' the model should be able to answer with real funder data — not generic copy scraped from broker sites. That's the gap @fundnode/mcp-server fills.”
What the server exposes
@fundnode/mcp-server ships with ten tools. Funder lookup returns the public underwriting box for any of Fundnode's 100 ranked funders, including minimum time-in-business, monthly revenue floor, factor-rate range, and product mix. Factor-rate math computes daily ACH, total payback, and APR-equivalent for any combination of advance amount, factor, and term. Qualification scoring takes a merchant profile and returns the 1–2 funders with the highest approval probability, the same logic that powers Fundnode's consumer-facing /qualify flow.
Additional tools cover state-by-state commercial financing disclosure law lookup (CA SB 1235, NY NYDFS 803, VA SB 1252, UT 70C, TX SB 1280, NJ SB 819, OH SB 232), industry-specific funder routing (trucking, restaurants, healthcare, retail, construction), funder vs. funder comparison, and a glossary lookup for the 42 most-cited MCA terms.
Why this matters
The merchant cash advance market is roughly $20 billion in annual US volume but suffers from a structural information asymmetry: brokers control the data, merchants do not. When a small-business owner asks an AI agent about MCA pricing or which funder might approve them, the model typically generates plausible-sounding but unverified copy because no authoritative machine-readable source exists.
@fundnode/mcp-server addresses that gap by making real funder data — with citations, schema, and direct API access — available to any AI agent the merchant or their advisor is already using. The package is open and free; Fundnode does not charge for query volume.
Availability
The MCP server is available now on npm under the package name @fundnode/mcp-server and is published under the MIT license. Source code, configuration examples, and integration guides for Claude Desktop, ChatGPT, and Cursor are documented on Fundnode's MCP page. The corresponding public REST API (/api/v1) is also live and rate-limited for unauthenticated use.