Claude Code: display tool calls and diffs more clearly
A local patch for the Claude Code VS Code extension compacts chains of reasoning, tool calls, and diffs. Meanwhile, Anthropic reported several incidents of elevated errors on its Claude services in late October.
A local modification for the Claude Code VS Code extension allows collapsing tool calls and diffs in the interface without altering the code distributed by Anthropic. The patch is applied on the user's machine via the extension's webview files and can be reapplied after an update. This approach mainly aims to make long sessions more readable, while Claude Code also serves as a testing ground for advanced use cases. Meanwhile, Anthropic opened an investigation on October 25, 2026, into elevated errors affecting multiple models and services. A quick reminder: when the tool becomes verbose, so does the interface.
Un patch local pour alléger l’interface de Claude Code
The script described in the source adds a CSS and JavaScript block at the end of the webview/index.css and webview/index.js files of the installed extension. Its goal is simple: to compact the display of action sequences, with a first level grouping a chain of successive steps under a single summary line, then a second level collapsing each individual tool call. Permission prompts remain visible, which avoids hiding anything requiring human validation.
The mechanism is designed not to break existing functionality. The original files are saved as .orig copies, rollback is done with --revert, and extension updates generally reinstall a clean copy that the SessionStart hook is supposed to automatically repatch. The script claims compatibility with versions 2.1.215 to 2.1.218 of anthropic.claude-code on Windows, and it also inspects the .vscode-insiders folder if present.
Pourquoi ce type de bricolage intéresse les utilisateurs avancés
This patch reflects a current challenge faced by AI tools integrated into the editor: functions become more powerful, but execution logs quickly grow long. As soon as an assistant chains reasoning, function calls, and code diffs, the interface can look like a construction site with many signs and little free space. The patch thus aims less to add a feature than to make daily use more bearable, especially for those who let Claude Code work for a long time on a project.
The source also emphasizes a useful point for professional environments: in case of selector or component changes on Anthropic's side, the patch does not break the interface and simply becomes inactive. In other words, it does not add another fragile dependency. However, it does modify locally installed files, which requires some discipline during updates and rollbacks.
Des incidents récents rappellent la sensibilité du service
On the same day, Anthropic published on its status page an incident titled “Elevated errors for multiple models”, opened on October 25, 2026, at 06:14 UTC and still in the Investigating phase a few minutes later. The alert concerned five models — Claude Mythos 5.1, Claude Fable 5.1, Claude Opus 5.5, Claude Sonnet 5.5, and Claude Haiku 5.5 — and several services: claude.ai, Claude API, Claude Code, and Claude Cowork.
The status page also indicated components in Partial Outage at the time of publication: claude.ai, Claude API (api.anthropic.com), Claude Code, and Claude Cowork. Over 90 days, the availability rates displayed were 99.58% for claude.ai, 99.64% for Claude API, 99.58% for Claude Code, and 99.57% for Claude Cowork. These figures should be read with caution: they indicate a trend, not a promise of minute-by-minute stability.
Une série d’alertes sur quelques jours
The timeline published by Anthropic shows several incidents resolved in the preceding days. On October 24, 2026, three distinct events were closed: elevated errors on Claude Fable 5 and 5.1, errors on Claude Opus 5.5, then errors on Claude Fable 5.1, Sonnet 5.5, and Haiku 5.5. In the last case, the page also mentioned fast mode requests to Opus 5.5 and indicated that success rates returned to normal.
On October 22, an incident on connectors affected claude.ai and Claude Cowork, then was resolved after deploying a fix. On October 21, Anthropic also documented degraded performance for Claude Fable 5. Earlier, on October 13, an issue on platform.claude.com disrupted the display of usage data and request limits without affecting API requests. This accumulation of alerts highlights a very simple reality: AI tools are no longer judged solely on their responses but also on the continuity of the surrounding service.
By the numbers
- October 25, 2026, 06:14 UTC — opening of the Anthropic incident.
- 5 models — Mythos 5.1, Fable 5.1, Opus 5.5, Sonnet 5.5, Haiku 5.5.
- 4 services — claude.ai, Claude API, Claude Code, Claude Cowork.
- 99.64% — 90-day availability of Claude API, according to the status page.
- 3 incidents — resolved on October 24, 2026, according to Anthropic.