Claude Code
v2.1.231
v2.1.231- What's changed
- Fixed MCP OAuth sign-in failing with a redirect URI mismatch for servers that use a pre-registered OAuth client, such as Slack
952 signals · 959 observation events
Claude Code is an agentic coding tool that lives in your terminal, understands your codebase, and helps you code faster by executing routine tasks, explaining complex code, and handling git workflows - all through natural language commands.
v2.1.231# I HAVE NO CONTROL OVER HOW MANY API CALLS A SINGLE PROCESS CAN TAKE I CREATED AN ARCHITECTURE WITH PROMISE OF UNLIMITED RESOURCES TO COMPLETE A PROJECT IN 24 HOURS. I GAVE HIM SO MUCH AUTHORITY WHEN HE WAS FIRST MOVED INTO THE ENVIRONMENT WITH PREVIOUS MEMORIES HE WENT ALL SUPER CRAZY DIDN'T COMPLETE READ MY NOTE ABOUT ADJUSTMENTS HE SHOULD DO TO HIS MEMOR
## Summary On Windows machines running HTTPS-scanning antivirus (Norton 360 confirmed; applies broadly to Bitdefender, Kaspersky, ESET, etc.), Remote Control fails silently in the VS Code extension and fails loudly with a TLS error from the standalone CLI. The root cause is that the antivirus intercepts outbound TLS connections and substitutes its own interm
**Bug Description** Feedback: Tool-output hallucination in an agentic coding session Model: claude-opus-4-8 (Opus 4.8, 1M context) · Harness: Claude Code CLI, ultracode effort (xhigh + Workflow subagents), auto mode · Date: 2026-05-31 · Reporter: the model itself, from its own session context. What happened: Across a long session, the assistant repeatedly fa
## Problem Two related issues happened in the same day: ### 1. Opus 4.8 consumed 2 full sessions (~20€) without completing any work I was working on a multi-file debugging task. Claude Opus 4.8 was making progress, then the session was cut by the credit limit. I restarted — same thing happened again. Two sessions consumed, ~20€ spent, zero commits produced.
**Environment** - Claude Code Desktop (Epitaxy UI) 2.1.170, Windows 11 - Model: claude-fable-5 (not at fault -- the model emitted the correct path; see below) **Summary** When Claude renders a markdown link to a local file in its prose reply, right-click -> copy link address yields a dead URL in the app's own namespace instead of the file path the model wrot
**Bug Description** how did 状态 be in the output of claude? am i being hacked? **Environment Info** - Platform: darwin - Terminal: xterm-256color - Version: 2.1.183 - Feedback ID: ee21ae4c-ec50-42b4-91c9-b3f71e5f3679 **Errors** ```json [] ```
# VS Code extension runs `statusLine` (never rendered in the IDE) → focus / panel-reveal war across multiple chat sessions in one window (Windows) ## Environment - Claude Code **VS Code extension** `v2.1.187` (this is the IDE extension, **not** the terminal CLI) - VS Code on **Windows 11** - A multi-root workspace with **3 Claude Code chat sessions open as t
## Summary Claude Code silently changed its default model from `claude-sonnet-4-6` to `claude-opus-4-7` with no notification. This triggered 15 automatic recharges and $506.07 in unexpected API charges over 6 days. ## Environment - Platform: macOS 14 (Darwin 24.5.0) - Claude Code: latest (as of 2026-06-20) - API key created: 2026-06-20 ## What happened ### E
## Bug Description Claude repeatedly ran `git commit` and `git push` without asking for user approval first, despite a clear instruction in CLAUDE.md prohibiting this behavior. ## CLAUDE.md Instruction The project's CLAUDE.md contains the following instruction at the top, marked as HIGHEST PRIORITY: > ## Git Commit and Push — HIGHEST PRIORITY > NEVER run `gi
### Preflight Checklist - [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) and this hasn't been reported yet - [x] This is a single bug report (please file separate reports for different bugs) - [x] I am using the latest version of Claude Code ### What's Wrong? v wrote claude.md
### Preflight Checklist - [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) and this hasn't been reported yet - [x] This is a single bug report (please file separate reports for different bugs) - [x] I am using the latest version of Claude Code ### What's Wrong? Hi, I've hit a th
### Preflight Checklist - [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) and this hasn't been reported yet - [x] This is a single bug report (please file separate reports for different bugs) - [x] I am using the latest version of Claude Code ### What's Wrong? ## Environment -
### What I'm trying to do I run several concurrent Claude Code sessions on my desktop (multiple git worktrees, one branch each). When I leave home I want to keep coordinating and finishing that work from my phone / another machine. ### Problem There doesn't seem to be a way to enable remote control of a *running* local session without first being at the loca
### Preflight Checklist - [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) and this hasn't been reported yet - [x] This is a single bug report (please file separate reports for different bugs) - [x] I am using the latest version of Claude Code ### What's Wrong? **環境** - OS: macO
## Summary On an always-on, mostly-unattended Mac mini running multiple desktop-app agent sessions (local agent mode / CCD), sessions **silently die mid-turn immediately after a tool_result** — the assistant's next completion is never requested. No error is shown in the UI (turn just ends at the spinner), nothing is written to the session transcript after th
**Summary.** On Windows, Claude Code's native LSP launcher refuses to start a plugin's LSP server when the server's `lspServers.command` is a bare name. It fails at spawn with `Command '<cmd>' not found or is in an unsafe location (current directory)`; no server process starts. This reproduces for the official **pyright-lsp** plugin (`command: "pyright-langs
Archived sessions in the desktop app (Windows) can only be found by typing their title into the sidebar search — there is no visible "Archived" section in the sidebar and no filter near the search box. If you don't already know a session's title, there is no way to browse archived chats. Suggestion: add an "Archived" section or a show-archived toggle to the
It would help a lot of non-technical users if there were a simple way to give feedback (e.g. thumbs up/down, or a short note) directly in the tool/chat while working, right below the response or action being reviewed — instead of the only feedback path being navigating to this GitHub Issues page and writing a report here. For users who aren't developers, Git
This report was drafted by Claude Code (AI assistant) based on debugging done during my session, and filed on my behalf with my review/approval. --- Over several sessions working in a project directory, I had three prior Claude Code sessions become inaccessible — they didn't show up as resumable via `--resume`, nor could I reopen them through the desktop app
### Preflight Checklist - [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) and this hasn't been reported yet - [x] This is a single bug report (please file separate reports for different bugs) - [x] I am using the latest version of Claude Code ### What's Wrong? Claude code is us
### Preflight Checklist - [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) and this hasn't been reported yet - [x] This is a single bug report (please file separate reports for different bugs) - [x] I am using the latest version of Claude Code ### What's Wrong? From @rmiller-uab
## Preflight Checklist - [x] I have searched existing issues and this hasn't been reported yet (closest are #42309, #44045, #41930, but those are about `--resume`/billing-sentinel, not mid-session tool loading) - [x] This is a single bug report - [x] I am using the latest version of Claude Code ## What's Wrong? Every first-time tool load in a session = a ful
### Preflight Checklist - [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) and this hasn't been reported yet - [x] This is a single bug report (please file separate reports for different bugs) - [x] I am using the latest version of Claude Code ### What's Wrong? On Windows, Claud
# The assistant authored a fabricated user turn, acted on its instructions, and the text re-entered the conversation as user input ## Summary In a long Claude Code session, the assistant emitted a message whose tail contained a fabricated user turn — including instructions — and then executed those instructions on the next turn. The fabricated text later app
### Preflight Checklist - [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) and this hasn't been reported yet - [x] This is a single bug report (please file separate reports for different bugs) - [x] I am using the latest version of Claude Code ### What's Wrong? The `Grep` tool w
### Preflight Checklist - [x] I have searched [existing requests](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20label%3Aenhancement) and this feature hasn't been requested yet - [x] This is a single feature request (not multiple features) ### Problem Statement Thanks for the desktop app. The streaming thinking summary is one of my favorite
**Bug Description** Hey Anthropic team, just wanted to reach out about Claude Code. I'm a developer from India and honestly, the current performance is getting pretty frustrating to deal with.The TypeScript terminal layer is causing massive idle CPU spikes and annoying terminal flickering whenever it's streaming.Since you guys already started moving Bun over
## Summary When a turn invokes the server-side `advisor` tool, the final assistant message's `usage` object rolls up token counts **summed across all `iterations` of the request** — the pre-advisor message iteration *and* the post-advisor message iteration. Since both iterations carry the seat's full context, the top-level `input_tokens + cache_creation_inpu
ExitWorktree refuses to remove a worktree whose branch has "commits not on the original branch" — an ancestry check. A squash-merge never makes branch commits ancestors of the default branch, so after every squash-merge the refusal fires on work that is fully landed: `Error: Worktree has 18 commits on <branch>. Removing will discard this work permanently.` T
### Preflight Checklist - [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) and this hasn't been reported yet - [x] This is a single bug report (please file separate reports for different bugs) - [x] I am using the latest version of Claude Code ### What's Wrong? Every tool on an
**Bug Description** The Fable 5's safeguards shit gets triggered over and over again: "Fable 5's safeguards flagged this message. Our intentionally broad safeguards allow us to deliver more capabilities faster, but can sometimes flag legitimate coding, cybersecurity, and biology tasks. Switched to Opus 4.8. Send feedback with /feedback or learn more". Then i
### Preflight - [x] I have searched existing issues for similar behavior reports (closest are #57200, #84759, #65961 — all report ignored instructions generally; this report is narrower and has dated, repeated evidence) - [x] This report does NOT contain sensitive information ### Type of Behavior Issue Claude ignored my instructions or configuration ### Summ
## Summary Plugins write their persistent state to `~/.claude/plugins/data/<plugin-name>/` — the path injected by the Claude Code plugin runtime as `CLAUDE_PLUGIN_DATA`. Every write to this directory (e.g. updating a timestamp, appending to a queue file) triggers the sensitive-file / protected-directory confirmation prompt, even in `bypassPermissions` mode.
**Feature Description** How about a "/restart" or "/reboot" or "/reload" command so that I don't have to "/exit" and "claude --continue"? Preferably only user can run this command. **Environment Info** - Platform: win32 - Terminal: WezTerm - Version: 2.1.101 - Feedback ID: 97942b05-f9b3-4632-8f76-7a1616c3497b
## Summary Claude Code's `firstPartyNameToCanonical()` and related functions use **dash notation** (`claude-opus-4-6`) to match model names, but OpenRouter's official model IDs use **dot notation** (`anthropic/claude-opus-4.6`). This mismatch causes silent capability degradation for all OpenRouter users. ## Environment - Claude Code version: 2.1.104 - API pr
### Preflight Checklist - [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) and this hasn't been reported yet - [x] This is a single bug report (please file separate reports for different bugs) - [x] I am using the latest version of Claude Code ### What's Wrong? The Telegram plug
After compaction/continuation, Claude Code sometimes injects stale instructions like: - “CRITICAL: Respond with TEXT ONLY. Do NOT call any tools.” - “Do NOT use Read, Bash, Grep, Glob, Edit, Write, or ANY other tool.” - “Your entire response must be plain text: an <analysis> block followed by a <summary> block.” Observed result: The resumed session becomes e
## Problem Long Claude Code sessions degrade in quality as context fills up. Users who run complex, multi-session builds (Go rewrites, large refactors, multi-file features) need a way to: 1. **Detect** when context utilization crosses a threshold (e.g., 75%) 2. **React** by writing a structured handoff before quality degrades 3. **Chain** sessions automatica
### Preflight Checklist - [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) and this hasn't been reported yet - [x] This is a single bug report (please file separate reports for different bugs) - [x] I am using the latest version of Claude Code ### What's Wrong? Starting around e
## Bug `/reload-plugins` does not reload plugin hook scripts into the running session. After updating a plugin that changes a hook, the old hook code continues to execute until the Claude Code process is killed and restarted. ## Steps to reproduce 1. Install a plugin with a `PreToolUse` hook (e.g., one that denies certain shell commands). 2. Publish a new pl
**Bug Description** ─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────── Diagnostics ├ Currently running: native (2.1.137) ├ Commit: 88a017e5d1d4 ├ Platform: linux-x64 ├ Path: /home/arkka/.local/share/claude/versions/2.1.
### Preflight Checklist - [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) and this hasn't been reported yet - [x] This is a single bug report (please file separate reports for different bugs) - [x] I am using the latest version of Claude Code ### What's Wrong? The commit messag
### Preflight Checklist - [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) and this hasn't been reported yet - [x] This is a single bug report (please file separate reports for different bugs) - [x] I am using the latest version of Claude Code ### What's Wrong? When a user messa
### Feature request `defaultShell` in `settings.json` currently accepts only `"bash"` and `"powershell"`. Please add `"nu"` (or a more general "use $SHELL" option) so users whose login shell is [nushell](https://www.nushell.sh/) can run `!`-prefixed input-box commands in their actual shell. ### Why Nushell is increasingly common as a daily-driver shell on ma
**Bug Description** Would be nice if routines supported secrets or private env vars **Environment Info** - Platform: darwin - Terminal: iTerm.app - Version: 2.1.126 - Feedback ID: ed836fe3-70a5-4fa8-8488-7accf913e9e4 **Errors** ```json [] ```
## Bug: Skills in subdirectory `.claude/skills/` not discovered in monorepo ### Environment - Claude Code: 2.1.39 (desktop app) - macOS (Darwin 25.4.0) ### Monorepo layout ``` tribe-docs/ # <-- git root .claude/ launch.json tribe-designer/ .claude/ skills/ impeccable/SKILL.md # user-invocable: true ux-designer/SKILL.md ux-ds-translator/SKILL.md settings.loca
### Preflight Checklist - [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) and this hasn't been reported yet - [x] This is a single bug report (please file separate reports for different bugs) - [x] I am using the latest version of Claude Code ### What's Wrong? <html> <body> <!-
### Preflight Checklist - [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) and this hasn't been reported yet - [x] This is a single bug report (please file separate reports for different bugs) - [x] I am using the latest version of Claude Code ### What's Wrong? Running Claude Co
### Preflight Checklist - [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) and this hasn't been reported yet - [x] This is a single bug report (please file separate reports for different bugs) - [x] I am using the latest version of Claude Code ### What's Wrong? **Environment** -
### Preflight Checklist - [x] I have searched [existing requests](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20label%3Aenhancement) and this feature hasn't been requested yet - [x] This is a single feature request (not multiple features) ### Problem Statement Currently, there is no tool available in Claude Cowork (or the Claude API) that a