Claude API Console’s New Dashboard and a Detailed Guide to Sonnet Extended Output

If you’ve recently been using the Claude API to build apps or agents, the most noticeable changes are in the console and documentation: the Claude API now has new usage and cost dashboards, release notes have been filled out, and Claude Sonnet 3.5 in the Claude API supports longer outputs. Below, I’ll explain the new features around what developers care about most: “see clearly, write longer, and track.”

Claude API output limit doubled: better delivery for long-form text and code

In the Claude API, Claude Sonnet 3.5’s maximum output tokens increased from 4096 to 8192, making it suitable for returning longer summaries, reports, code generation, and multi-step reasoning results in a single response. To enable extended output in the Claude API, you need to add the specified beta request header in your request: "anthropic-beta": "max-tokens-3-5-sonnet-2024-07-15".

In practice, it’s recommended to reserve “long output” in the Claude API for scenarios that truly need it—for example: generating a complete PRD, outputting a runnable script, or delivering an annotated refactoring plan in one go. For requests that only need a conclusion, it’s still better in the Claude API to control max_tokens to avoid burning your budget on verbose wording.

Claude API Console adds usage and cost dashboards: more transparent billing

With the new “Usage” and “Cost” tabs in the developer console, Claude API consumption can finally be viewed by USD amount, token count, and API key. For teams that share a Claude API key across multiple environments and projects, this makes it faster to pinpoint anomalous calls and the source of sudden cost spikes.

Even more practical is that you can track experimental calls and production calls separately in the Claude API: split API keys across services, then compare token and cost curves in the dashboard. This way, Claude API budget discussions no longer rely on “gut feeling”—you can speak directly with data.

Claude API release notes are live: no more guessing updates

Anthropic has added more complete release notes in the documentation, covering updates across the Claude API, the Claude Console, and the Claude app. For developers who depend on stability, this reduces the risk of “the interface changed and I didn’t know,” which is especially important when you use the Claude API in critical paths (customer support, review workflows, code generation).

It’s recommended to make Claude API release notes part of your routine checks: before upgrading models or adjusting prompt templates, confirm whether there are behavioral changes, limit adjustments, or new beta capabilities. In the long run, this saves far more time than troubleshooting production issues after the fact.

Claude API docs and Cookbook expansion: more like a hands-on manual

Beyond feature updates, the Claude API documentation has also been redesigned, and the Claude Cookbook has been expanded with new guides around core skills like “citations, retrieval-augmented generation (RAG), and classification.” For anyone integrating the Claude API with a knowledge base, ticketing system, or content moderation pipeline, this content is closer to real engineering implementation rather than just concepts.

If you’re building RAG right now, it’s recommended to first use the Cookbook to lock in the “citation format” and “constraints on retrieved results,” then return to the Claude API side to tune parameters. This makes outputs more controllable and also makes it easier to present explainability and perform quality acceptance in your product.