May 2026 Summaries

Lago has developed the Agent SDK for Python and TypeScript to streamline the token usage tracking and billing process for AI features leveraging large language models (LLMs). This SDK addresses the common issue faced by teams building AI applications, where they must repeatedly construct complex middleware to parse token usage from various LLM providers, which often change formats and cause disruptions. By wrapping the LLM client with the Lago Agent SDK, developers can automatically attribute token usage to customers, normalize data across different providers, and send this information to Lago for billing. This reduces the burden of maintaining separate middleware for each provider, allows for consistent customer billing, and integrates seamlessly with existing infrastructure such as caching proxies and AI gateways. While the SDK currently handles token extraction and attribution, Lago plans to introduce a pricing table feature that will automatically update provider pricing and apply user-defined markups, further simplifying the billing process for both AI-native teams and B2B SaaS companies integrating AI features.

Anthropic's recent pricing changes for its AI software, Claude, have sparked controversy by highlighting a shift in how AI usage is billed, moving from a user-centric model to one that distinguishes between human and agentic usage. Previously, AI subscriptions like Claude's bundled costs in a way that felt seamless to users, who were unaware of the underlying metering, but the introduction of distinct credit pools for agentic usage has revealed the true costs associated with automated processes that can operate continuously without human intervention. This shift has caused backlash among developers who built workflows around the previous pricing model, leading to a perception of a price increase despite the company's assertion that subscription limits remain unchanged for interactive use. As Anthropic grapples with these changes, competitors like OpenAI's Codex are capitalizing on the situation by offering easier adoption paths without immediate metering concerns, although they too will eventually confront similar challenges as AI products evolve. The broader industry trend suggests a move towards more transparent billing systems that cater to both human and automated uses, requiring companies to find ways to make these changes feel beneficial rather than punitive to users.