Context window & memory limits — page 3
Issues 401–600 of the 6,285 matching this topic. The trend chart and topic overview are on page 1.
- [BUG] VS Code extension crashes with OOM error: conversation history accumulates to 700MB+ causing infinite crash loop
- Feature Request: Auto-compact warning Hook
- [BUG] Claude Code crashes due to OOM
- [BUG] Max 20x upgrade not reflected in weekly limits — depleting at Max 5x rate or worse (upgraded July 16, 2026)
- [FEATURE] Let the agent trigger /compact itself (agent-invokable compaction)
- [BUG] CLAUDE_CODE_DISABLE_1M_CONTEXT ignored in 2.1.156
- [Regression] advisor() still breaks sessions post-compaction via parentUuid tree mismatch -- root cause identified, workaround script included
- [Feature Request] Remove or make optional resume summary/compaction prompt on session resume
- Feature Request: Hook for assistant text output (PreResponse / PostResponse)
- [BUG] Autocompact triggers at 17% on 1M context model (effectiveWindow likely capped at 200k)
- Allow tools/skills to programmatically clear context and inject a continuation prompt
- Feature: configurable auto-compaction threshold
- Feature Request: Native token compression for CLAUDE.md and instruction files
- [BUG] /heapdump fails with EEXIST on Windows
- [BUG] Bug: Teammates do NOT load skills despite documentation stating they do
- [BUG] Error: Error during compaction: Error: Conversation too long. Press esc twice to go up a few messages and try again.
- autoCompact: false ignored - compaction triggers at ~78%
- [FEATURE] Enable Task tool for subagents with configurable depth limit
- Context window tracking in statusline doesn't reset on /clear
- [BUG] Could not process image
- [Bug] Instant Auto-Compact increases token consumption even when disabled
- Compaction fails with 400 error when thinking blocks present in assistant message
- [FEATURE REQUEST] Better compaction is possible without any fancy machinery, MCPs, and with very little plumbing.
- [BUG] SDK (TypeScript at least) fails to invoke hooks
- Conversation Compaction Failure: Unexpected Error When Attempting to Reduce Context Length
- Configurable Compaction Warning Threshold to Prevent Conversation Degradation
- [BUG] Bash tool is completely unresponsive - no commands are executing or returning output
- Tool call intermittently emitted as literal "count" (antml: prefix dropped, "could not be parsed") in long Opus 4.8 1M-context sessions
- [BUG] Claude Code spawned ~3,000 parallel bash processes in 17 seconds, causing OOM crash and hard reboot.
- [FEATURE] Surface MCP tool outputSchema to the model context, not just the renderer
- [BUG] Model picker shows Opus 4.7 (200K) but session silently runs in 1M context — burned 18% of 5-hour Max limit on a small prompt
- System prompt size grew ~70K tokens between v2.1.89 and v2.1.96, making sessions unusable without frequent manual /compact
- Status line: show current effort/reasoning level
- Add 'compact context' option when accepting plans in Plan Mode
- [Feature Request] Add opt-in toggle for 1M context in Cowork — Max users lost access after feature flag rollback
- [FEATURE] Make ToolSearch activity visible in the UI like other tool calls
- feat: expose rate limit data in statusline JSON
- Opus 4.6 update implies 1M context but Claude Code hard-caps at 200K — causes session crashes (Max 20x)
- [FEATURE] Remove skills from context when not in use.
- Bug: /context command executes /compact instead
- [FEATURE] VS Code Extension: Add "Clear Context" option when exiting Plan Mode
- Token counting ~2x inflated in 2.1.7 vs 2.1.2
- [Bug] Compaction fails with "Conversation too long" error, leaving session unrecoverable
- [BUG] # Critical Bug: Background Bash Processes Cause Infinite System-Reminders and Token Exhaustion
- Hooks broken again in v2.0.31 (regression after v2.0.30 fix)
- API Request Size Exceeds Context Window Limit
- Claude Code says conversation is to long and auto compact did not work
- [FEATURE] Make the auto-memory MEMORY.md index size limit (200 lines / 25KB) configurable
- [BUG] compact
- [Bug] Auto-Compact Not Triggered
- Allow per-subagent exclusion of CLAUDE.md / .claude/rules (parity with built-in Explore/Plan)
- [BUG] CLI 2.1.154 builds invalid request after context ops (compact/clear/model-switch): system role at messages[0], and modified signed thinking blocks → 400 Error -> Wedged
- [BUG] Claude Code uses entire 5 hour token limit in one prompt with zero feedbacks or results
- [FEATURE] Support custom statusLine in VS Code extension (parity with CLI)
- [Bug] Auto-compact triggers far below 1M window on Opus 4.7 [1M] — severe token burn regression on Max 5x (post-2.1.116)
- [Bug] Session/weekly token limits deplete disproportionately fast with Opus orchestrator + Haiku subagents
- [Feature Request] Do not include slash command output in conversation context (/release-notes takes 10% of context alone).
- Context window detection fails for third-party Anthropic-compatible providers
- Cloud MCP servers (Gmail, Google Calendar) auto-injected without user consent — no opt-out mechanism
- [Bug] v2.1.88 yanked release caused wasted session + significant token loss
- Korean (CJK) characters corrupted to U+FFFD in Claude Code responses on macOS + VSCode
- Expose background task count in statusLine JSON input
- Compact prompt should preserve historical context across multiple compactions
- [FEATURE] Rolling asynchronous compaction
- [FEATURE] Extend deferred tool loading to built-in subagents
- Feature Request: Deferred Loading for Skills (SkillSearch, like ToolSearch)
- Auto-compact should trigger earlier / threshold should be configurable
- Opus 4.6 1M context model unusable on Max plan - immediate rate limit error
- [FEATURE] Expose Plan Quota/Usage Data to StatusLine
- Bug: Memory leak causing OOM kill on Linux (15-17GB memory consumption)
- [BUG] /context command header shows 0k tokens while breakdown shows ~108k tokens
- [Bug] Context limit reached prematurely before /context indicates 20% remaining
- Context display shows 15% free but reports 'Context limit reached'
- [BUG] Claude Code VS Code extension leaks worker processes when creating new chats
- [Feature Request] Make file read token limit dynamic based on model capabilities
- /compact fails with 'tools: Tool names must be unique' error
- [DOCS] Missing PermissionRequest hook details in Hooks Guide and Input Reference schema
- [BUG] Autocompacting Loop Causing Massive Token Usage Spikes OCT 15 2025
- Reduce token waste from redundant file change notifications
- [Bug] Auto-Compact Not Triggering Correctly at 0% Context Remaining
- [BUG] Conversation Compaction Corrupts Tool Use/Result Pairs
- [FEATURE] Ability to selectively remove context
- # [CRITICAL] Auto-compact triggers at 8-12% context making Claude Code unusable
- Context Window Exhaustion During Conversation Compaction
- Feature Request: Hierarchical Context Management (Conversation Stack)
- [BUG] Claude Code suddenly dumb like sh**
- Anthropic API Context Limit Exceeded During Compaction
- Question: Does Claude include CLAUDE.md in its own context after compact?
- Keyboard Interrupt Handling Breaks Text Input Context
- Context hit 1M token limit with no auto-compaction and no recovery path
- [FEATURE] Selective session retention: allow pinning/bookmarking important sessions to prevent auto-cleanup deletion
- Context/compact button triggers compact on single click with no confirmation or undo
- [Bug] Anthropic API Error: Cannot modify thinking blocks in assistant message
- False Usage Limit and Request blocked
- [BUG] CLI 2.1.145 sends literal "claude-opus-4-7[1m]" as model name on session resume, hits 404, silently falls back to 200K
- [BUG] Bun 1.3.14 JSC GC segfault (SlotVisitor::drain) on AMD Ryzen 5 5500 / Linux x86_64_baseline — all platform binaries identical
- [BUG] Skill subagent inherits parent's 1M-context tier but not extra-usage entitlement
- [Feature Request] Allow programmatic/agent-initiated compaction
- [BUG] Max plan: context window stuck at 200K instead of 1M (hasExtraUsageEnabled not updated after Pro → Max upgrade)
- Bug: --resume loads 0% context on v2.1.91 — three regressions in session loading pipeline (source code verified)
- [BUG] High Token Burn Due to Redundant Context Resubmission & Compaction Loops
- [Bug] CLI hangs or stalls during execution, performance significantly degraded
- Feature Request: Hook or plugin to deduplicate/evict data from context window
- [BUG] Claude Code Cannot Handle API Requests >5 Minutes Long
- claude.ai Skills and Cowork plugins silently injected into Claude Code context — no opt-out, ~6k tokens wasted per session
- Memory leak: ArrayBuffers consuming 4+ GB within seconds of startup
- [Bug] Context amnesia in long sessions — constraints silently dropped as context grows
- [FEATURE] Option to disable built-in statusline content when using custom statusline
- [Bug] Context limit reached prematurely despite 70%+ free context available
- [BUG] huge memory leak
- Premature "Context limit reached" at ~51% when submitting image during active processing
- STOP DUMBING DOWN YOUR MODELS WHEN YOU ARE MAKING UPGRADES OR CHANGES! THIS IS DANGEROUS...AT LEAST LET THE USERS KNOW!
- [Bug] Autocomplete selection defaults to /compact instead of /context match
- [DOCS] Undocumented "Clear Context" transition options in Plan Mode
- Feature Request: Add context usage percentage to hooks and enable automatic /compact execution
- [BUG] claude is losing chat history
- [BUG] JavaScript heap exhaustion (OOM crash) during heavy Task tool/agent usage
- Bug: Context becomes full and /compact stops working after using subagents (Task tool)
- 💖 Community Learnings: 7 Critical Token-Wasting Patterns (700K+ tokens saved!)
- conversation compression
- [FEATURE] Allow model selection in slash command frontmatter
- [Bug] Conversation Length Exceeds Compaction Limit
- [BUG] CRITICAL: Account-Specific Auto-Compact Corruption - Complete Service Loss for Max Plan Subscribers
- Context Memory Loss After Auto Compact Summary
- Compact Summary Generation Failure in Conversation Summarization
- Context Warning Disappears Before Reaching Zero in Version 0.2.107
- Claude Code Prompt Guidelines Enforcement (Feature)
- Assistant text in the same message as an AskUserQuestion/ExitPlanMode tool call is silently dropped (not rendered, not persisted to session JSONL)
- [FEATURE] /nudge — send an ephemeral one-turn instruction that doesn't persist in conversation context
- [FEATURE] VS Code extension: show model, Ultracode/mode, effort, and usage indicators next to the chat input (desktop app parity)
- Graceful handling when hitting usage limits mid-session
- Allow reverting to previous Windows Desktop UI or add theme/layout settings
- [BUG] Max 20x plan ($200/mo) — quota exhausted in ~1h, making the product unusable for our team
- [BUG] Aggressive clearing of conversation history at ~2k tokens on Opus 4.6 (1M context) v2.1.89
- [BUG] Session resume loads zero conversation history — silently drops all context
- [FEATURE] Invert plan mode "clear context" setting — show by default, hide via opt-out
- Unrecoverable session after dropped tool_result — rewind/restore/summarize all fail
- Expose usage/rate limit data via CLI flag or hook event
- [FEATURE] Structural File Reading — AST-aware Read tool for large files (80% context reduction measured)
- [FEATURE] Show active loops (`/loop`) in the status bar and task list
- [BUG] Memory leak: RSS grows continuously (~1GB/min) with zero user interaction after session start
- [FEATURE] Skill profiles: role-based scoping of active skills
- [BUG] 2.1.63 silently defaults to 1M context model without UI indication, causing excessive usage consumption and a BILLING issue.
- [BUG] Context compaction loses critical working knowledge mid-session
- [FEATURE] Context usage threshold hook event with plan-and-continue workflow
- [BUG] Compaction death spiral - 211 compactions consuming all tokens with zero progress
- Subagent results silently overflow context, causing unrecoverable session crash
- Feature: Lazy-load skills and MCP servers (name-only stubs until invoked)
- Claude Code crashes with SIGABRT when session JSONL files exceed V8 heap limit
- [BUG] Subsequent skill calls duplicate content in model context
- Status line doesn't refresh context tokens after /clear
- [Bug] Status line token counts not resetting after /clear or /compact commands
- [FEATURE] Please bring back SessionEnd output logs!
- [BUG] Skills truncated at 30 makes remaining skills undiscoverable
- [Bug] Output token limit settings not being applied, agent output capped at 32k tokens
- [BUG] Context window shows 45k autocompact buffer on fresh session with no conversation history
- [FEATURE] /model command in the SDK
- [FEATURE REQUEST] hook on session init
- Auto Compact Blocking Claude Interaction: Continuous Compaction Loop
- Context persistence across sessions - major workflow disruption for complex projects
- Error compacting conversation Context left until auto-compact: 0%
- Node.js Memory Allocation Failure in Latest Version
- Expose permission_mode in the statusline payload + statusLine.hidePermissionModeIndicator (refresh of #46419/#47345)
- [FEATURE] Toggleable columns in agent view: model, effort, and context usage
- [BUG] Agent Teams: Single teammate spawn creates 10-151 duplicate worker instances, each consuming full context and actively editing files
- Opus 4.7 ([1m] variant) capped at 200K context in Claude Code Desktop on Max plan
- [BUG] CLAUDE_AUTOCOMPACT_PCT_OVERRIDE=50 does not trigger autocompact — context reaches 100% requiring manual /compact
- [Bug] Extended Thinking causes 7–8 minute stalls with no output and silent token consumption
- Image dimension error blocks ALL responses instead of gracefully degrading
- Auto-compaction triggers at ~6% context usage on Opus 4.6 (1M context)
- [FEATURE] Open-source Claude Code under Apache 2.0
- Feature request: SessionEnd / BeforeSessionEnd hook event
- [BUG] Team agent teammates don't inherit [1m] context window variant from leader (v2.1.80)
- [BUG] VSCode Extension shows 200k context window inspite of 1M
- "Simplest approach first" system instruction causes quality degradation in complex projects
- [Bug] Opus 4.6 default model alias not using 1M context window
- Session name reverts after context compaction
- [Feature] Graceful context compaction — PreCompact hook data + background compaction option
- [Feature Request] ExitPlanMode lacks defer/abandon option, trapping users in plan mode
- [FEATURE] Expose partial compaction as /compact --from parameter for non-interactive use
- TaskOutput burst polling causes unrecoverable 'Prompt is too long' when multiple background agents complete simultaneously
- [Bug] Context indicator wraps character-by-character in narrow terminals
- [Bug] Context usage percentage inconsistency between statusline data and /context command
- [FEATURE] Expose session/weekly rate limits to statusline script JSON input
- [BUG] Maximum call stack exceeded
- [Bug] Context leak with background subagents causing unexpected context limit crash
- [Bug] Auto-Compact triggers prematurely with reduced context window
- [Bug] Sonnet 4.5 context window fills up prematurely with large files
- [Bug] Token Compaction Error: "Error: Conversation too long" at only 97k tokens used??
- [BUG] /context slash command misreporting context usage as mostly free when in fact it is mostly used
- Claude AI usage limit prematurely reached
- Feature Request: Automatic Context Restoration After Autocompaction
- Auto Compaction Failure in SDK Mode with JSON Streaming Conversation Continuation
- [FEATURE REQUEST] - Initialisation prompt to use after compacting
- [BUG] Claude keeps forgetting about the context in CLAUDE.local.md after compacting
- [FEATURE] Index repositories for faster lookups
- [BUG] Long pause then "interrupted by user" for grep and git commands
- Per-subagent `advisor` frontmatter field (pair each subagent's executor with its own advisor)
- [Feature Request] --resume should restore the last active model (including [1m] variants)
- [FEATURE] Exempt CLAUDE.md from compaction