Kelviq enters a market where the traditional subscription model is being upended by the variable costs of large language models. For years, software-as-a-service was defined by the per-seat license, a predictable transaction with high margins. The rise of AI agents and API-driven services has changed the math. When a company's cost of goods sold fluctuates with every token generated or every agentic loop completed, a flat monthly fee becomes a liability. Kelviq addresses this by combining two distinct financial layers: the Merchant of Record model and a real-time usage metering engine.

Based in San Francisco, the company was started by a team that previously managed monetization at Branch, a mobile infrastructure firm that reached $100M in annual recurring revenue. That background is visible in Kelviq’s architectural choices. They prioritize billing debt reduction, a term they use to describe the technical overhead of building custom logic for trial periods, feature gating, and usage caps. Instead of treating billing as a peripheral database entry, Kelviq treats it as a real-time state machine that dictates what a user can actually do within an application.

The Merchant of Record distinction

Most developers start with Stripe, which provides the plumbing for moving money but leaves the legal and tax burden on the merchant. For a global AI startup, this means registering for VAT in the European Union, GST in India, and sales tax across dozens of US states. Kelviq is a Merchant of Record, meaning they technically buy the service from the developer and resell it to the end user. They take on the liability for tax collection and remittance, compliance, and invoicing. This model is common among established players like Paddle, but Kelviq focuses on high-frequency usage tracking specifically for modern AI stacks.

Metering and entitlements

The technical core of the platform is a usage metering system that handles the reportUsage calls familiar to developers using cloud infrastructure. In an AI context, this allows a developer to track token consumption or successful agent completions in real-time without building a separate infrastructure for counting and aggregation. This data feeds directly into the billing engine, which can then enforce entitlements. These are the logic gates that determine if a user has exceeded their credit limit or if they have access to a specific model version.

By linking the billing plan directly to the user state, Kelviq removes the lag between payment and access. If a customer upgrades their plan to access a more capable agent, the feature unlocks immediately because the billing system and the application’s permission layer are synchronized. This is particularly relevant for autonomous agents that might need to scale their resource consumption based on the complexity of a task.

Competitive position

Kelviq competes with incumbent MoRs like Paddle, which has a larger footprint but often feels tailored to traditional software downloads or simple subscriptions. It also competes indirectly with metering-only services like Orb or Lago. The tradeoff Kelviq offers is simplicity through consolidation. By putting tax, payments, and usage metering in one bucket, they aim to be the default choice for small AI teams that would rather spend engineering hours on model performance than on tax code.