Context window & memory limits
Context-length overflows, compaction issues, and heap/out-of-memory crashes.
6285 related issues · 908 open · 5377 resolved · peaked at 908 open (2026-08)
Backlog trend over time
The line shows how many issues in this topic were open at the end of each month, derived by
counting each issue from its creation date until it was closed. First-party data from the
anthropics/claude-code tracker.
All context window & memory limits issues
Ranked by community engagement (reactions and comments).
- Allow MCP tools to be available only to subagent
- CC v2.1.100+ inflates cache_creation by ~20K tokens vs v2.1.98 — same payload, server-side
- [BUG] [URGENT!!!] Claude Code is hanging / freezing / stuck on heaps of prompts for 5-20minutes or more.
- [BUG] Pro Max 5x Quota Exhausted in 1.5 Hours Despite Moderate Usage
- [FEATURE] Allow setting plan naming scheme per-repo
- [FEATURE] VSCode Extension: Display context usage percentage in UI
- Feature Request: Enable 1M Context Window for Claude Code
- [Feature Request] Claude Desktop: Option to execute commands in WSL instead of Windows
- Claude.ai MCP servers auto-injected into Claude Code without opt-in — causes OOM crashes on resource-constrained systems
- Messages disappear from terminal display during conversation after v2.1.89 update
- [BUG] Error during compaction
- [FEATURE] Allow slash command usage over remote-control
- [Feature Request] Enhanced /compact with file-backed summaries and selective restoration
- [BUG] Auto-compact not triggering on Claude.ai (web & desktop) despite being marked as fixed
- [BUG] Critical memory regression in 2.1.27 - OOM crash on simple input
- [BUG] CLI output formatting artifacts break copy/paste - workarounds waste tokens
- Claude Code Memory Leak - Process Grows to 120+ GB RAM and Gets OOM Killed
- [FEATURE] Auto-continue after subscription rate limit resets
- Slash commands not supported in /remote-control UI
- [BUG] No working mechanism to review previous context after compaction, plan-mode clear, or branch navigation — data preserved but UI inaccessible
- [BUG] Extremely rapid token consumption. There appears to be a critical issue with token calculation and quota management. Users, including myself, are experiencing an unusually high rate of token consumption
- Critical Memory Leak: Claude Code Consumed 129GB RAM and Caused System Freeze
- Sibling tool call errored: parallel tool calls cascade-fail when one fails
- Feature request: Native Zellij support as alternative to tmux for Agent Teams and /terminal-setup
- [BUG] Claude Code defaults to 1M context on fresh session with no workaround on Pro plan
- [BUG] Abnormal / inflated rate limit / session usage
- Expose rate limit usage (session %, weekly %) to statusLine configuration
- [Feature Request] Add 'claude model list' CLI command for non-interactive model queries
- [FEATURE] Add user-configurable keyboard shortcut actions
- Feature Request: CLI Commands for MCP Server Enable/Disable (Hook Automation Support)
- [FEATURE] Add status bar to Desktop App
- [Bug] Potential session/cache leakage between workspace instances or consumer accounts
- [FEATURE] PostCompact Hook Event and Compaction Content Control
- [BUG] Scheduled tasks cannot access MCP connectors until a user message warms the session
- Recurring crashes: JavaScript Heap Out of Memory while 'thinking'
- [BUG] Fable 5 prompts 'usage credits required' on Max plan from 2026-07-20, the day it became standard on Max (saved model claude-fable-5[1m])
- [FEATURE] Add --no-auto-compact command switch to Disable the auto-compaction
- [BUG] resume flag fails to maintain conversation context after hitting usage/context limits
- [FEATURE] Smart Session Rename
- [BUG] Statusline context_window JSON contains cumulative tokens instead of current context usage
- [BUG] Conversation compaction intermittently corrupts conversation state: “Compaction failed unexpectedly” and silent prompt submission failures
- [FEATURE] Skill invocation tracking and usage analytics
- [Bug] Auto-Compact Erases Entire Chat History Without Warning
- [CRITICAL] Context auto-compact stuck at 0% - Tool Unusable
- [BUG] API Error: 500 {"type":"error","error":{"type":"api_error","message":"Internal server error"}}
- [BUG] Cowork (Not Claude Code) 1M context window unavailable on Max 5x — regression from March 20
- Feature Request: Persistent Memory Across Context Compactions (59 compactions, built our own)
- [BUG] Error loop between 'context low' and '/compact'
- [BUG] Latest version aggressively hides all previous output during active session
- [BUG] Memory leak in long-running idle Claude Code sessions
- Allow curl
- Add dynamic loading/unloading of MCP servers during active sessions
- [BUG] Error: Error during compaction: API Error: Usage credits required for 1M context · turn on usage credits at claude.ai/settings/usage, or use --model to switch to standard context
- [Bug] Plan mode resurfaces old plans after session compaction with Claude 4.5 Opus
- [FEATURE] Expose rate limit/session usage data to statusLine configuration
- [Bug] Anthropic API Error: Usage credits required for 1M context window with opus-plan model
- [BUG] [1M] Sonnet 1M context window disappeared after Claude Code update to v2.1.45 — regression on Max plan ($200/mo)
- [Bug] /context command reports 200K max for Claude Opus 4.6 instead of 1M
- [FEATURE] Expose token usage and cost data in hook inputs
- [BUG] claude starts and 3 seconds later exits.
- [FEATURE] Support slash commands (/clear, /compact) from Channels (Telegram/Discord)
- [BUG] Claude forgets everything in CLAUDE.md after compaction
- [BUG] Edit tool silently converts tabs to spaces, causing repeated match failures on tab-indented files
- [BUG] "Request too large (max 20MB)" error on small files (99KB PNG)
- /compact fails with 'Conversation too long' when context limit is reached
- Feature: Enter/resume existing worktrees, configurable branch naming, hook removal control
- Rename the conversation
- [BUG] VS Code Extension has an incomplete slash command list compared to the CLI version
- Both /compact command and auto-compaction broken - 'conversation too long' error
- [CRITICAL] Subagent spawning and subagent pattern bugs trigger infinite recursion, infinite token usage, grossly inefficient token usage, and lost accumulated subagent work.
- [FEATURE] VSCode extension auto-includes opened files as context — should be opt-in by default
- [BUG] Extreme token consumption — quota depleted in minutes with normal usage
- Subagents fail with 'prompt is too long' when user has many MCP servers (tool definitions exceed 200k)
- [Bug] Auto-compact never triggers despite statusline reporting "100% context used" (v2.1.153, Max sub, 200K mode)
- [FEATURE] Provide way to exclude parent CLAUDE.md files from auto-loading in subdirectories
- Opus 4.6 (1M context) hits 'API Error: Rate limit reached' on Max subscription
- Add --max-context flag to cap context window usage
- [Feature Request] Subagents should inherit parent agent context to avoid redundant codebase scanning
- [BUG] Auto compact does not trigger at 100% context window and Claude Code stops itself on cli
- Sonnet 4.6 1M context unavailable on Max 20x plan - only Opus 1M appears in /model
- [Feature Request] Session Checkpointing and Branching
- Fable 5 model behavior: 16 min of silent tool calls, then AskUserQuestion about never-shared findings during incident response
- [BUG] Context limit reached · /compact or /clear to continue
- [FEATURE] Add support for 1M Context on Claude Sonnet 4.5
- [BUG] Skill `model:` frontmatter triggers 429 "Extra usage required" on 2.1.76 when session uses opus[1m]
- Feature Request: Add Context Pruning as Alternative to Compacting
- Feature request: Session persistence & reconnection for Remote Control server mode
- [BUG] Conversation compaction intermittently fails or deadlocks chats (claude.ai web/mobile, post–Jan 15 regression)
- JSON Parsing Error: Invalid Low Surrogate in Request Body
- Feature Request: Runtime MCP Server Toggle
- [BUG] No postCompact hook
- Feature Request: Restore Time & Token Indicators + Add Context Usage Display
- API Request Failures: Persistent 529 Overloaded Errors
- Opus 4.6 shows 200K context instead of 1M on Claude Max plan (v2.1.75)
- [BUG] /rate-limit-options command repeatedly auto-executes after session reset, causing excessive token consumption
- Feature Request: 1M Context with Max Plans in CC
- [BUG] Terminal laggy scrolling and jumping with long history
- [FEATURE] Claude-invocable conditional /compact for automated workflows
- [BUG] Claude Code consistently degrades in performance over long sessions
- [Feature Request] Expose how much context is left before auto-compact
- [BUG] API Error: Connection closed mid-response — occurs immediately in new context window
- Skills consume full token count at startup instead of progressive disclosure (frontmatter only)
- SessionStart hooks not working for new conversations
- [BUG] /compact causes Claude Code to ignore CLAUDE.md
- Feature Request: Show active skill in status line
- [BUG] Keep hitting "Request size exceeds model context window"
- Cowork: Context compaction permanently removes scrollable conversation history from UI
- Opus 4.6 comprehensive regression: loops, memory loss, ignored instructions - daily professional user report
- [FEATURE] Tool result transform hook for content sanitization
- Model fabricates tool output (and even a user instruction) when a parallel batch is partially cancelled / appears empty
- [FEATURE] Bring 1M context window to Max plan subscribers
- [BUG] Opus 4.6 with [1m] context window returns "long context beta not available" on Max subscription
- [BUG] slash command auto-complete is broken (highlighted command doesn't match what I'm typing)
- [FEATURE] Allow Claude to self-initiate context compaction
- [BUG] Turning auto-compact OFF won't allow to continue until you fill the 200K tokens window anymore (it worked before).
- Anthropic API 529 Overloaded Error During Compaction Process
- [FEATURE] Indexed transcript references in compaction summaries for surgical context recovery
- Prompt bar UI bug in v2.1.23: text disappears, lines compact
- [FEATURE] Expose Agent Context in Hook Event Payloads for Multi-Agent Observability
- Include effort level in statusLine hook input
- [BUG] Token Limit Hard-Stop Without Warning or Auto-Compaction
- [BUG] Terminal clears frequently and for no reason, hiding conversation history
- [FEATURE] Claude clear context for itself
- Chat Branching: Spawn side-chain conversations with selective merging back to main
- [BUG] Compact Conversation Infinite Loop
- [Bug] /compact and auto-compact fail with 'Tool names must be unique' when MCP servers are configured
- [BUG] Agent-team teammate pane crashes with "<Box> can't be nested inside <Text>" when ※ recap renders inline markdown code spans
- [FEATURE] Add setting to configure default option in plan mode approval dialog
- Expose token usage data to statusline scripts
- [BUG]
- Feature Request: Pro Plan 5-Hour Window Usage Visibility
- [Feature Request] Configurable autocompact context %
- [BUG] Prompt is too long (Claude for VS Code)
- [BUG] API Error: 400. Each `tool_use` block must have a corresponding `tool_result` block in the next
- Rewind history lost after compaction in sessions with many rewinds (branching tree)
- [Bug] Context limit reached with 32k tokens free space and auto-compact disabled
- [BUG] Plan mode broken - "Auto mode is unavailable for your plan" blocks all input
- [BUG] Silent context degradation — tool results cleared without notification on 1M context sessions this issue documents three separate mechanisms (microcompact, cached microcompact, session memory compact)
- [BUG] /context command no longer display in chat window on VS Code ext version 2.1.45
- [FEATURE] Configurable Auto-Compact Threshold in VS Code Extension
- VSCode extension v2.1.141: "Unhandled case: [object Object]" banner from assertNever on unknown streaming-event types
- Opus 1M context on Max plan requires extra usage despite docs saying 'included with subscription'
- Context low warning and /compact failure despite having 86% free context space
- [BUG] Claude Desktop Drops/Reverts Long Multi-Step Tasks Without Warning
- Fable 5 → API-only: please reconsider a subscription path (👍 + comment here if it affects you)
- [PROPOSAL] Expose compact/session lifecycle hooks for external memory layers
- [FEATURE] Ask questions for compaction
- [Bug] File loading adds 70% token overhead due to line number formatting
- [BUG] CLAUDE.md instructions completely ignored after context compaction
- [BUG] Appeal Form Redirect Loop After Account Restriction
- [FEATURE] Add /fork (conversation branching) support to VS Code extension
- Compaction fails with 'Conversation too long' at 48% context usage (Opus 4.6)
- Add a setting option to force re-reading of CLAUDE.md after /compact
- VS Code extension: add a verbose toggle equivalent to CLI Ctrl+O
- Terminal scrollback wiped during session (alternate screen buffer cleared on redraw)
- [BUG] API Error: Rate limit reached incorrect in CLI
- Feature Request: Expose rate limit data in statusline JSON input
- [FEATURE] Expose /usage and /context data in statusLine JSON
- Feature: Expose session ID as $CLAUDE_SESSION_ID environment variable
- Compaction Failure: Conversation Length Exceeds Limit
- [A11y feature request] Built-in option to speak Claude's responses aloud
- Explore agent always fails with "Prompt is too long" when many MCP tools are connected
- Feature Request: Expose Max/Pro plan usage limits for status line display
- [BUG] Context percentage mismatch between /context, statusline API, and internal limit trigger
- Auto-compact fails with Opus 4.5 extended thinking blocks
- A session cannot determine whether its auto-memory index loaded whole, truncated, or not at all
- API Error: Stream idle timeout - partial response received
- [BUG] Context window defaulting back to 200k
- [BUG] 1M context window removed from Desktop Code tab model picker after v1.1.7714 update on Max plan
- [Feature Request] Include actual context window usage in statusline JSON
- [BUG] Instant auto-compaction not instant - compaction takes same time as before v2.0.64
- Feature: Long-context (≈1M) support for Claude Sonnet 4 in Claude Code
- Opus 4.8 intermittently emits malformed tool calls: stray `count`/`call` token, then tool call without `antml:` prefix rendered as plain text
- Max plan: 1M context window shows 200k, Remote Control unavailable
- [BUG] Claude Desktop: Send button silently fails - message disappears when context is high
- Remote Control regression: server-injected `<system-reminder>Message sent at …>` prefix breaks slash commands from mobile/web
- [FEATURE] Claude Desktop - Archive individual conversations from the sidebar
- Agent team lost when lead's context gets compacted during long session
- Feature Request: Add PreCompact and PostCompact hooks for custom context management
- [Bug] VSCode Extension: User-level rules (~/.claude/rules/) not loading into context
- [Bug] Anthropic API Error: Usage credits required for 1M context window
- [BUG] Session name lost after context compaction
- Session permanently stuck on 'Rate limit reached' after transient rate limit
- Compaction clears screen - can't scroll up to see previous context
- [Bug] Manual Compaction Fails with "Conversation Too Long" Error
- [BUG] workings with agents makes the terminal super slow and hangs often
- [FEATURE] Visible context window usage indicator in Cowork
- [Feature] Subagent's TaskOutput should return final result text, not full JSONL log
- [Bug] Compaction Fails Unexpectedly with "Conversation Too Long" Error
- [Feature Request] Migrate Claude Code to Toad Terminal UI Framework
- New Hook suggestion: PostCompact
- [FEATURE] UserInputChange hook — expose the prompt input buffer to hooks, mirroring MessageDisplay
- KRITISCH - ZWEITER KATASTROPHALER DATENVERLUST in 4 Tagen - 1 Monat Arbeit PERMANENT ZERSTÖRT - Zahlender Kunde fordert SOFORTIGE Maßnahmen
- [BUG] After compaction the terminal resets and i cannot scroll up and see what all were discussed, changed and done.
- Feature Request: Improve Claude Code Token Management with MCP Servers
- [MODEL] Claude Code weaponizes user's own CLAUDE.md rules to fabricate context warnings and manipulate user into reducing workload
- [Critical] Max Plan subscriber: Opus 4.6 instruction-following regression breaks production workflows — 24-hook enforcement system cannot compensate for model-level degradation
- Opus/Sonnet 1M context variants broken on Max 20x — misleading 'Rate limit reached' error
- [BUG] Auto compact off, but asking me to compact before 100% context usage
- [FEATURE] Native context visibility for self-regulating multi-context workflows