Claude Code v2.1.122
Released Apr 28, 2026 (4 months ago).
45 issues reference this build
1 still open
44 resolved (98%)
"References this build" means the issue body mentions running 2.1.122 (from the
bug-report template). It reflects report volume, not a proven defect count — an issue may mention a
version without being caused by it. First-party data from the anthropics/claude-code tracker.
Top issues reported on v2.1.122
Ranked by community engagement (reactions and comments).
- [BUG] MCP OAuth: Claude doesn't auto-refresh access tokens, daily "Connection expired" despite valid refresh token
- [BUG] Write permission allowlist ignores relative file_path that resolves to absolute-path rule target (still broken in 2.1.122 — see #29098 / #37220 / #38391)
- Oversized images in resumed conversations should auto-downscale instead of hard-blocking the turn
- [BUG] Remote routines lose MCP connector permissions every run — tools blocked in ephemeral container
- [BUG] Github Connect button in Organization Settings doesn't do anything
- [Feature Request] Add toggle keybinding to show running background agents with live output
- [BUG] System prompt constant qD4 unconditionally injects security-adjacent text into every session, priming the API cyber classifier
- [Bug] Claude Code fails to generate coherent research questions from case study analysis
- Skill tool feedback loop: getPromptForCommand body re-injected as isMeta after every model-invoked Skill call
- [BUG] AskUserQuestion returns 'undefined is not an object (evaluating H.map)' on subsequent calls within same turn (2.1.122)
- [BUG] Claude Code OAuth breaks when workspace setting "Allow creating new API keys in default workspace" is disabled
- CC tasks disappear from TUI after /compact (session ID not preserved)
- [BUG] /doctor v2.1.122 mislabels home-scope .mcp.json parse error path with cwd-template, masking actual file location
- MCP server registered via 'claude mcp add' not surfacing tools in conversational interface when claude.ai web connectors are also present
- [BUG] MCP deferred tool activation — client tool list doesn't refresh after notifications/tools/list_changed
- [BUG] Plan Mode - Proposed plan on local machine does not match proposed plan on remote machine
- claude mcp add --scope user writes to a location claude mcp list does not read
- [Bug] Persistent session feedback prompt cannot be dismissed
- [Bug] /mcp command produces no dialog or output
- [Bug] Claude silently marks tasks complete despite unresolved blockers instead of surfacing failures
- Project-root `config/` directory auto-deleted when Codex plugin installed (data loss)
- [Bug] /ultrareview quota consumed on failed runs: empty findings with worktree confusion and result-posting crash
- permission allowlist over-escapes Bash() patterns containing parens/quotes/dollars, producing 'unhandled node type: string' prompts mid-session
- [Feature Request] Add autonomous execution mode for automated workflows
- Support custom sampling parameters (temperature, top_p, top_k) for third-party models via ANTHROPIC_BASE_URL
- [Bug] Parallel sub-agents inherit full parent context, causing excessive token consumption without cost warnings
- [Bug] Unexpected spike in API usage for tier 5 user
- Claude Code 2.1.122 crashed with embedded Bun v1.3.14 segfault at address 0x38
- [Bug] Hidden messages prevent visibility of Claude responses and tool prompts
- [BUG] Failed ultrareview - lost my free attempts
Changelog
What's changed
- Added
ANTHROPIC_BEDROCK_SERVICE_TIERenvironment variable to select a Bedrock service tier (default,flex, orpriority), sent as theX-Amzn-Bedrock-Service-Tierheader - Pasting a PR URL into the
/resumesearch box now finds the session that created that PR (GitHub, GitHub Enterprise, GitLab, and Bitbucket) /mcpnow shows claude.ai connectors hidden by a manually-added server with the same URL, with a hint to remove the duplicate- Clarified the
/mcpmessage shown when an MCP server is still unauthorized after the browser sign-in flow - OpenTelemetry: numeric attributes on
api_request/api_errorlog events are now emitted as numbers, not strings - OpenTelemetry: added
claude_code.at_mentionlog event for@-mention resolution - Fixed
/branchproducing forks that fail with "tool_use ids were found without tool_result blocks" when the source session contained entries from rewound timelines - Fixed
/modelnot showing the Effort option for Bedrock application inference profile ARNs, and those ARNs not receivingoutput_config.effort - Fixed Vertex AI / Bedrock returning
invalid_request_error: output_config: Extra inputs are not permittedon session-title generation and other structured-output queries - Fixed Vertex AI
count_tokensendpoint returning 400 errors for users behind proxy gateways - Fixed
spinnerTipsOverride.excludeDefaultnot suppressing the time-based spinner tips - Fixed ToolSearch missing MCP tools that connected after session start in nonblocking mode
- Fixed
!exit/!quitin bash mode terminating the CLI instead of running as a shell command - Fixed images sent to newer models being resized to 2576px per side instead of the correct 2000px maximum
- Fixed remote control session idle status redrawing twice per second, which could flood
tmux -CCcontrol pipes and pause the terminal - Fixed assistant messages appearing blank in some sessions due to a stale view preference
- Fixed a malformed hooks entry in
settings.jsonno longer invalidating the entire file - Voice mode: keybindings bound to Caps Lock now show an error since terminals don't deliver Caps Lock as a key event