Settle's MCP Integration Pioneers Agentic Procure-to-Pay for CPG Brands

NEW YORK – October 29, 2026 – The journey from an innovative software prototype to a profitable, market-dominating platform often hinges on a company's ability to play nicely with others. In the complex world of business-to-business financial technology, closed ecosystems are rapidly becoming a relic of the past. Settle, a cash flow management and working capital platform tailored for inventory-based businesses, has recognized this shift, announcing the launch of its public API and a Model Context Protocol (MCP) server.

This strategic release opens Settle's procure-to-pay platform directly to data warehouses, custom software, and, crucially, artificial intelligence assistants like Anthropic's Claude. By doing so, Settle is laying the groundwork for what the industry is dubbing "agentic procure-to-pay"—a framework where AI does not just summarize data, but actively handles the operational busywork across a company's disparate systems.

Backed by heavyweight investors including Ribbit Capital, Kleiner Perkins, and Founders Fund, Settle has already processed over $8 billion in payments and originated more than $4 billion in working capital loans. However, this latest technological milestone represents a critical pivot from being a standalone financial dashboard to becoming the invisible, intelligent engine powering the back office of Consumer Packaged Goods (CPG) brands.

"Every business we talk to is rebuilding how they work with AI, and the companies moving fastest want their tools to talk to each other," said Alek Koenig, CEO and founder of Settle. "We built Settle's API and MCP server so a brand's AI can work alongside our platform and their other systems, not around them. This is the foundation for agentic procure-to-pay: software that keeps purchasing, payables, and cash flow moving on its own, with our team behind it."

The Rise of Agentic FinTech via Anthropic's MCP

To understand the commercial significance of Settle's announcement, one must look at the underlying technology: the Model Context Protocol. Anthropic's MCP is an open standard designed to allow AI models to securely and effectively interact with external tools and data sources. While large language models (LLMs) have historically been confined to text generation and data summarization, MCP acts as a standardized bridge, allowing these models to read live balance sheets, query payables, and orchestrate procurement actions.

Settle's adoption of MCP marks one of the earliest public implementations of this standard within the B2B fintech and supply chain payments sector. While competitors like Brex and Ramp have heavily invested in AI for expense categorization, receipt matching, and anomaly detection, their approaches have largely relied on proprietary AI integrations within their own walled gardens. Similarly, legacy giants like Tipalti and Bill.com utilize AI for invoice processing and fraud detection, but have not yet fully embraced open protocols that allow external, autonomous AI agents to orchestrate workflows across the entire enterprise stack.

By adopting MCP, Settle is betting that the future of corporate finance software is interoperable. Instead of forcing a CPG brand to log into the Settle dashboard to check a vendor payment, a finance director can simply ask their internal AI assistant, "What do we owe our suppliers next month?" The AI, securely connected via the MCP server, queries Settle's payables data alongside the company's bank accounts and returns an accurate answer in seconds. This shift from reactive software to proactive, agentic AI represents a massive leap in commercial viability, positioning Settle as a foundational layer rather than just another application.

Taming the Fragmented Tech Stack of CPGs

For CPG brands and inventory-heavy businesses, the operational back office is notoriously fragmented. A typical mid-market e-commerce brand operates on a chaotic patchwork of software: Shopify for storefront sales, a traditional ERP for resource planning, separate accounting software like QuickBooks or NetSuite, disconnected 3PL (Third-Party Logistics) systems for warehouse management, and an endless array of Excel spreadsheets bridging the gaps.

This fragmentation creates a massive commercial bottleneck. Reconciling inventory levels against incoming purchase orders and outgoing vendor payments requires hundreds of hours of manual data entry. Cash flow forecasting becomes an exercise in historical guesswork rather than real-time analysis.

Settle's API and MCP server are specifically designed to untangle this web. By allowing developers and AI agents to tap directly into Settle's payables and working capital tools, finance teams can automate the reconciliation process. Early adopters of the platform are already leveraging the API to build custom cash flow dashboards and deploy autonomous operational agents that monitor inventory levels in a 3PL and automatically draft purchase orders within Settle when stock runs low.

Industry analysts note that solving this specific pain point is highly lucrative. Software that can demonstrably reduce manual reconciliation hours and improve cash flow visibility has a significantly higher retention rate and lifetime value. By embedding its capabilities into the broader tech stack, Settle makes its platform stickier and much harder for a growing CPG brand to rip out and replace.

The Safety Dilemma: Who Approves the Wire?

However, the transition from prototype to profit in the realm of agentic finance is fraught with regulatory and security challenges. Allowing autonomous AI agents access to corporate payables and, more critically, working capital drawdowns introduces a complex safety dilemma.

The core risk lies in the "black box" nature of some AI models. If an AI agent hallucinates a vendor invoice or misinterprets a data point from a 3PL system, it could theoretically initiate an unauthorized payment or draw down thousands of dollars in high-interest working capital without human oversight. Furthermore, AI agents must navigate stringent compliance frameworks, including Anti-Money Laundering (AML) regulations and sanctions screening, which traditionally require strict human governance.

Corporate treasury experts are quick to point out that while the automation of data aggregation is highly desired, the autonomous execution of money movement remains a red line for many enterprise controllers. To commercialize this technology successfully, Settle and its peers must implement ironclad safeguards.

In a production environment, this requires strict Role-Based Access Control (RBAC). An AI agent tasked with cash forecasting might be granted read-only access to Settle's endpoints, while an agent managing procurement might have write access to create purchase orders, but lack the final authorization to release funds. Implementing a "Human-in-the-Loop" (HITL) architecture—where the AI does 99 percent of the prep work but a human controller clicks the final "approve" button on a wire transfer—is essential. Additionally, comprehensive audit trails that log every query and action taken by the AI are mandatory for maintaining SOC 2 compliance and satisfying internal auditors.

Navigating the Future of Autonomous Finance

The commercialization of AI in B2B fintech is no longer about adding a chatbot to a user interface; it is about fundamentally rewiring how corporate data flows and how financial decisions are executed. Settle's launch of its public API and MCP server is a calculated move to capture the growing demand for interoperability and automation in the CPG sector.

As the company expands what AI agents can see and do across procurement, payables, and working capital, its success will depend on balancing the promise of frictionless automation with the uncompromising reality of financial risk management. If Settle can prove that its agentic integrations are not only efficient but mathematically and operationally secure, it will set a new standard for how inventory-based businesses manage their capital. In the race to dominate the back office, the winners will be those who allow their software to fade into the background, letting intelligent agents do the heavy lifting.