Claude WorkbenchがConsoleに統合——Claude Code・Cursor開発者の移行チェックリスト【2026年8月】Anthropic will shut down Claude Workbench on August 17, 2026, merging its…

Anthropic has announced that it will retire Claude Workbench, the prompt-testing environment built into claude.ai, on August 17, 2026, and fold its equivalent capabilities into Claude Console at console.anthropic.com. The change matters to anyone who has built a workflow around iterating on prompts inside Workbench, particularly developers who pair Claude with coding tools such as Claude Code or the Cursor editor, because it requires action before the cutoff date to avoid interruption.

Claude Workbench has served as a lightweight space for composing, running, and refining prompts, letting users test how the model responds to different system prompts, parameters, and inputs before wiring them into an application. Console, by contrast, is Anthropic's broader developer hub, where teams manage API keys, monitor usage and billing, and reach documentation. Consolidating the two means prompt testing will now sit alongside the account and API management tools that production developers already use.

For developers, the practical concern is continuity. Prompts, saved configurations, or snippets created in Workbench should be reviewed and, where necessary, exported or recreated in Console ahead of the shutdown, since Anthropic appears to be positioning Console as the single destination for this functionality. Anyone who has bookmarked Workbench URLs, embedded them in internal documentation, or referenced them in onboarding material will likely need to update those links to point at console.anthropic.com.

The move is especially relevant to users of Claude Code, Anthropic's command-line coding assistant, and Cursor, a popular AI-assisted code editor that can call Claude models through the API. Both tools rely on the same underlying API credentials managed in Console, so centralizing prompt testing there is consistent with how these workflows already authenticate. Developers who tune prompts in Workbench and then deploy them through Claude Code or Cursor should confirm that their API keys, rate limits, and organization settings remain intact after migrating.

Prompt testing, sometimes called prompt engineering or prompt iteration, is the process of refining the instructions given to a language model to produce more reliable output. A dedicated interface for this task reduces the friction of editing code just to try a new phrasing, and it is a common feature across the major model providers. Keeping that capability close to billing, logs, and key management is a reasonable fit for teams moving from experimentation toward deployment.

The consolidation also reflects a broader industry pattern in which vendors separate consumer-facing chat products from developer platforms. claude.ai is Anthropic's general-purpose chat interface, while Console targets programmatic access; housing prompt testing in Workbench somewhat blurred that line. Merging it into Console arguably simplifies the product surface, though it also means casual users of claude.ai who occasionally experimented with Workbench will need a Console account to keep doing so.

Because the source material frames this as a migration checklist, the prudent steps for teams are straightforward. Inventory any Workbench-based assets now, verify Console access for everyone who needs it, re-test critical prompts in the new environment, and update any documentation or automation that references the old location. Completing this work before August 17, 2026 leaves room to catch discrepancies in how prompts behave across the two interfaces, which can occur even when the underlying models are identical.

Anthropic has not, based on the available information, indicated that pricing or model access will change as a result of the migration; the announcement centers on where prompt testing takes place rather than what it costs. That distinction is worth keeping in mind, as consolidations of this kind sometimes coincide with feature changes, but nothing in the current details points to that here.

For context, Anthropic has steadily expanded Console and its developer-facing surface as Claude adoption has grown among engineering teams, and tools like Cursor have made API-driven access to frontier models a routine part of day-to-day coding. Centralizing prompt testing within Console fits that trajectory. Developers should nonetheless watch Anthropic's official channels for detailed migration guidance as the date approaches, since specifics about data export, saved-prompt portability, and feature parity will ultimately determine how smooth the transition is in practice. Acting early, rather than waiting for the deadline, remains the safest way to avoid workflow disruption.