Anthropic rolls out AI spend-visibility tools as Claude Code posts 900% net revenue retention

Anthropic launches spend-visibility reporting for enterprises

Anthropic has rolled out a suite of reporting tools designed to show enterprise customers exactly what their token spending buys, TechCrunch reports. Announced Wednesday, the tools let managers inspect a team's output over a chosen date range — what was produced, what it cost and where workflows broke down — using a combination of internal analytics and employee surveys. The reports also surface recommendations for improving results, and Anthropic says the underlying data will not be used to train its models by default. The company is adding pooled group budgets as well, giving teams a shared allocation for managing token consumption.

The launch is aimed squarely at return on investment, which TechCrunch describes as a major concern among Anthropic's largest customers. While usage at venture-backed firms is what is driving token spend, the outlet notes that most of the Fortune 500 have yet to see major gains from deploying AI.

Financials point to Claude Code's traction

The product news landed in the same week that Anthropic shared financial figures with investors, which TechCrunch reports as evidence that Claude Code has become the commercial engine the frontier lab wanted. The company recorded an operating profit in the second quarter and posted net revenue retention of 900%, meaning customers spending with Anthropic in the third quarter of 2025 were spending nine times as much a year later.

Third-party usage data supports that picture. Jellyfish, a company that builds tools for software developers, published a report earlier this week based on agentic coding tool usage across 276,000 of its customers. Claude Code and other tools built on Anthropic models remain the most widely used in that sample, accounting for nearly a third of all proposed code changes submitted for review. Spending on those tools has grown roughly tenfold since the start of the year and now equals about 2% of total engineering cost.

Jellyfish chief scientist Nicholas Arcolano told TechCrunch that even if 2% sounds small, tenfold growth in under a year is the kind of line item most CFOs never planned for. His view is that the industry has reached the point where simply spending more is no longer enough — companies now have to optimize the spend and make it work.

Competition and lock-in

The reporting tools also arrive amid questions about competitive pressure. OpenAI's Astra model, released this summer, looked like a potential threat to Anthropic, but TechCrunch reports it has not slowed the company's adoption numbers so far. Nikolas Albarran, AI product lead at Jellyfish, told TechCrunch that Codex took a long time to show up in the firm's data at all, and that while chatter on social platforms suggests a heated race, Anthropic moved early with strong enterprise sales motions. Teams then put real effort into standardizing workflows around its tools, he said, and have not found a big enough reason to switch.

That stickiness is partly structural. Once an organization has built workflows around one vendor's agentic tools, switching carries a real cost, because each platform handles agents and files slightly differently. Albarran also pointed to the investment every major AI lab is making in forward-deployed engineering — teams of AI consultants embedded inside customer organizations to help them get results. Anthropic's new reports can be read as a lighter-weight version of that hands-on support, delivered through software rather than staff.

Why it matters

Anthropic is converting a usage boom into a retention story: an operating profit and 900% net revenue retention suggest Claude Code is working as a business, not just a research showcase. But that same growth creates its own problem — enterprise AI budgets are expanding faster than most companies can justify them, and reporting tools that measure output per token directly answer the question CFOs are now asking. Those tools do double duty: the more visibility customers gain into optimized workflows built on Claude, the more expensive it becomes to leave. Expect measurement and optimization, not just model quality, to become the next competitive front among AI vendors.