Context window & memory limits — page 2
Issues 201–400 of the 6,285 matching this topic. The trend chart and topic overview are on page 1.
- Feature Request: Agent-Scoped MCP Servers to Prevent Context Window Overflow
- [BUG] API Error: 400 Input is too long for requested model.
- [BUG] v2.1.117 embedded ugrep wrapper amplifies regex backtracking from grep-process-OOM into V8-heap-OOM (8 GB ceiling) — host freezes on WSL2
- [Bug] Sharp increase in token consumption since April 5-6 weekend — Max plan 5h limits hit in ~1-2h
- [BUG] Claude Code with Bedrock API often hits 4096 max output token limit on moderate complex tasks
- [Feature-Request] Allow calling (cahining) commands within prompt
- [FEATURE] Expose Claude.ai plan usage in statusLine JSON payload
- [FEATURE] Use # prefix for skills to distinguish from / commands
- Nerd Font Unicode characters not rendered in Claude Code UI (statusline, chat, file views)
- Token usage limit reached significantly earlier than in previous versions
- [Bug] /compact fails with "Tool names must be unique" even after disabling all MCP servers (v2.0.70)
- [BUG] Context compaction erases .claude/project-context.md instructions
- [FEATURE] Built-in status line with rate limit usage bars
- [BUG] Bug: Tmux scrollback buffer cleared during autocompact/compact
- [BUG] Context low warning shows 7% when actually 36% free (auto-compact disabled)
- How to get session ID from interactive session?
- Bundled ugrep OOMs the host: -E with two bounded {0,N} intervals explodes DFA construction to multiple GB
- [BUG] Usage credits required for 1M context - Pro plan blocked despite 17% usage
- Allow opt-out of built-in deferred tools via settings to reduce baseline context
- [Bug] System events delivered as user-role messages cause model to fabricate user consent and act on it
- [Bug] API Error: Rate limit reached despite 0% usage
- MAJOR BUG: Claude Code Stuck in Infinite Compaction Loop Making it Basically Useless!
- Auto-compact to avoid "input length and max_tokens exceed context limit" errors
- Feature request: auto-compact on idle timeout to prevent cache expiry cost
- [FEATURE] View full conversation history in UI
- [BUG] /context executes /compact instead, unless I ask nicely :-)
- [Bug] Anthropic API: Unexpected Weekly Token Limit Depletion
- Feature request: save session summaries from /compact
- [FEATURE] Show the actual model of each spawned subagent in the task/progress UI
- Opus 4.6 (1M context): autocompact triggers at ~76K tokens — 92% of context window wasted
- [Bug] TaskOutput returns raw JSONL transcript instead of subagent summary (regression in 2.0.77)
- Feature Request: Advanced Agent Orchestration with Hierarchical Delegation and Persistent Agents
- Feature Request: Scoped Context Passing for Subagents
- "Infinite Auto-Compact Loop Causing Performance Degradation"
- [BUG] Opus 4.7 [1M] context capped at 500K on Max x20 — server-side enforcement persists across fresh install (org_level_disabled stale flag)
- advisor() tool inflates reported input tokens by forwarding full transcript, triggering premature auto-compaction on extended context models
- [FEATURE] Per-model configurable autocompact threshold
- [BUG] Version 2.1.20 hides tool usage details (what files are being read, what search patterns are being used) behind Ctrl + O
- [BUG] Conversation compaction loses entire history, ctrl+e unresponsive
- Orphaned processes after terminal close cause runaway memory consumption (34GB+ per instance)
- Image Dimension Exceeds Max Size in Multi-Image API Request
- [BUG] /claude-api bundled skill saturates context unconditionally — ~77% spike from neutral question
- 93% rate limit consumed in ~3 minutes on fresh session (Max plan, v2.1.85)
- Claude Opus 4.6 1M context: self-reported degradation starting at 40%, recommending restart by 48%
- [BUG] Opus 1M context window suddenly returns "Rate limit reached" on both Max and Team plans (200K works fine)
- [FEATURE] Plan Lifecycle Hooks
- [BUG] Input Lag / Slow Typing Echo When Context is Full (Regression from 2.0.x)
- Context fills extremely fast, forced compact at ~70% instead of 99%
- 🚨 BUG: Persistent Unnecessarily Auto-Compact Triggered in an INFINITE LOOP - MAKING CLAUDE CODE USELESS!
- Custom model via ANTHROPIC_BASE_URL: no way to declare a context window >200k (status line & auto-compaction assume 200k)
- [Bug] Tool-use block markup intermittently corrupted with stray token prefix
- [BUG] Claude Desktop install causes Windows 10 Task Manager to crash when switching to More details
- [BUG] App intermittently goes permanently blank/white during active session — no crash logged, no recovery
- Make task tool reminder interval configurable
- [BIG FEATURE] Rework Plan Mode into a 1st-Class Plan System (Plan Tools + Draft Mode + Plan TUI)
- [BUG] Thinking mode 400 error on FIRST tool call in brand new conversation (Opus 4.5)
- [BUG] VSCode extension incorrectly reports 100% context
- [BUG] Session becomes completely unrecoverable due to thinking block API error
- [FEATURE] Resumable Subagent Sessions (Persistent Multi-Turn Context via Session ID)
- Having multiple MCP servers running Eats into Context Window
- [BUG] /compact fails with cyber safeguards false positive during legitimate defensive security session
- [Bug] OPUS 4.7 IS DOWNGRADED JOKE
- Opus 4.6 (Max $100) — Quota reaches 93% after minimal prompting in a single session
- [BUG] Terminal scrollback cleared mid-session without compaction (v2.1.74, iTerm2)
- Scroll back + Color bleaching + Memory leak fix -- npm i -g claudefix --- Cons: Plugs myself and runs an ad, makes claude immutable so it stays locked --- Slowly adding more festures to ennhance UX, youre a bot if youre bitching about safety in the comments.. Linux only
- [BUG] $100 subscription is getting capped super fast compared to the experience that I had in the past 4-5 months.
- /compact command fails with 'Conversation too long' error at 82% token capacity
- v2.1.92 regression: autocompact threshold reduced to 400k on Opus 4.6 (1M context)
- Expose reasoning effort level in status line JSON input
- Need ability to disable auto-memory (MEMORY.md)
- [BUG] /compact always fails with "Conversation too long" since v2.1.23 (regression)
- [BUG] Excessive scrolling in long terminal conversations
- [FEATURE] Pre-Compact Auto-Save and Improved Summarization
- [BUG] Haiku 4.5 incorrectly limited to 8192 output tokens instead of 64K
- [FEATURE] compact conversation tool
- Search for Python virtualenv in .venv/
- [FEATURE] Add /fork (conversation branching) support to VS Code extension
- Post-mortem 2026-04-28: 12 multi-agent coordination bugs surfaced across a single autonomous-overnight cycle
- [BUG] 2.1.111 introduced ~14% context window bloat at session startup (8% → 22%)
- [BUG] Task tool `model` parameter returns 404 - Blocks cost-optimized workflows (10-30x unnecessary spending)
- Memory leak causing V8 OOM crashes (SIGABRT) on extended sessions
- [Bug Report] Unable to compress conversation
- Feature Request: Expose plan usage/subscription limits in status line input
- [Bug] Background subagent tool calls exposed in parent context window
- [Bug] Compact Fails When Context Window Nearly Full, Requires Restart
- [ENHANCEMENT] MCP Resources Pagination Support for ListMcpResourcesTool
- Stuck retry loop: `400 thinking blocks cannot be modified` on large interleaved-thinking turns using AskUserQuestion
- [BUG] Claude stops responding mid-task after successful tool calls
- [BUG] [1M] Opus 4.6 1M context unavailable on Max 5x — works on another machine with same account
- Feature Request: Layered memory system for persistent cross-session context
- [BUG] /resume only shows ~5-10 most recent sessions, older sessions invisible despite existing on disk
- [Bug] /clear command fails to reset session token buffer
- :rocket: Introducing tweakcc - customize your Claude Code installation
- Compaction Failure: Conversation Length Exceeds Processing Limit
- Bundled claude-api skill injects entire 570KB SKILL.md (~210k tokens) into context on load
- Max plan: sonnet[1m] requires extra usage but opus[1m] is included - inconsistent
- [BUG] It auto interrups when doing any task
- Feature Request: Include timestamps in conversation messages visible to the model
- Feature request: lossless context cleanup before auto-compaction
- Permission dialog: add mid-granularity live controls (directory/session scoping)
- [Bug] Auto-complete incorrectly selects "compact" option when typing "/context"
- C:/Program Files/Git/model command persists globally, affecting separate terminal sessions
- [BUG] Sonnet 1M context window unavailable on Max 20x subscription - regression from ~2 weeks ago
- [Bug] Session token limit reached prematurely after upgrade to v2.0.64
- Feature request: /export-all that includes the whole content of a session, and is not limited by compaction or terminal buffer limit
- [Bug] Anthropic API Error: Usage credits required for 1M context with 1M context disabled
- [FEATURE] Claude desktop app / Claude Code - allow questions to be temporarily hidden so the user can read the context
- [FEATURE] Raise, remove, or make configurable the 10,000-char limit that persistHookOutput enforces on hook additionalContext (introduced in v2.1.89)
- API Error 500 (Internal Server Error) during active session
- [FEATURE] Support disabling claude.ai MCP connectors at the project level
- [BUG] transcript spontaneously disappears
- [BUG] 2 PERCENT OF BUDGET USED BY ASKING ABOUT BUDGET USE
- Image processing failure silently passes oversized images into context, bricking sessions
- [BUG] Agent Teams teammates lose [1m] context window suffix — root cause: tmux spawn hardcodes model without variant
- [BUG] In-process team agents lack the Agent tool (cannot spawn subagents)
- [FEATURE] Request: Model awareness of elapsed time between messages
- Context compaction fails with 'Conversation too long' when context limit is reached
- Cascading failure: auto-compaction, /compact, and rewind all fail at context limit — requires Ctrl+C to escape
- [BUG] "Context limit reached" triggers at 27% usage (68.9% free space) — regression in 2.1.33
- [FEATURE] Allow existing Claude Code sessions to join a team as teammates
- [FEATURE] Option to dim/inline the bypass permissions indicator in custom statusline setups
- [BUG] /context command output lost colors after v2.1.20
- Feature request: CLI flags/settings to hide default slash commands and agents from menu
- [Bug] Chat display issues with excessive spacing and missing content after recent updates
- [BUG] undefined is not a constructor (evaluating 'new G($,I,L)')
- [FEATURE REQUEST]: Enable configuration of Auto-Compaction
- Context window status bar shows 200k for Claude Sonnet 4.6, but model supports 1M tokens
- [FEATURE] Lazy context loading: extend the ToolSearch pattern to all context components
- Feature request: compactToolOutput setting to collapse tool call details by default
- Max plan Extra Usage charges during usage reporting outage — 1M context billing change undocumented
- Expose plan usage quota in statusLine JSON
- [BUG] Claude Code stating session limits hit and rejecting requests despite token quota being available
- Regression: Opus 4.5 "Shadow Downgrade" Significant reasoning quality drop & aggressive context compaction
- [Bug] Effective context window reduced from 200k to 160k tokens
- [BUG] Hundreds of lines of modified file notes being injected into context on EVERY user message via <system-reminder>
- [Feature Request] Plan mode should support plan files outside ~/.claude/plans directory
- [BUG] 400 Input is too long for requested model with `global.anthropic.claude-sonnet-4-5-20250929-v1:0`
- [FEATURE REQUEST]: Selective Context Compaction
- [BUG] Thinking block summaries truncated mid-sentence (showThinkingSummaries: true)
- Auto-compact stopped working for third-party API providers since v2.1.161
- Tool outputs expanded by default — no setting to collapse, inconsistent across machines
- [BUG] API Error: Extra usage is required for 1M context when invoking skill with model: frontmatter from Opus 4.6 [1m] session on Max plan
- [BUG] v2.1.87 idle process grows ~500 MB/min, OOM-killed within minutes on 16GB WSL2 system
- [BUG] No way to view pre-compaction conversation history — Ctrl+O/Ctrl+E broken, thinking blocks hidden
- Add weekly and session usage percentages to status line JSON input
- API stream hangs mid-response on complex generation tasks (all models: Opus, Sonnet, Haiku)
- [FEATURE] /alias command
- [FEATURE] Let me turn off Autocompact buffer and free 22.5% of context
- [Bug] Premature message compaction at low token usage threshold
- [BUG] 2.0.13 MASSIVE TOKEN USAGE DUE TO SYSTEM REMINDER
- [FEATUE REQUEST] 200K is not enough. Please bring 1M token.
- Remote Control: built-in slash commands (/clear, /context) not intercepted — delivered to the model as plain text
- [FEATURE] Visible context-window usage indicator in Cowork (parity with Codex / Claude Code status line)
- [Bug] /compact fails with "Usage credits required for 1M context" even after /model set to Sonnet 4.6
- Feature: Fine-grained compaction controls (partial compaction, boundary-aware, compaction context)
- Bad System prompt effects
- /login shows "Login successful" but immediately reverts to "Not logged in"
- feat: MCP server push notifications (unsolicited messages to client)
- [FEATURE] Add setting to disable context window compacting reservation
- Expose plan usage limits in statusline JSON input
- Feature Request: Expose conversation context metrics to status line hooks
- [BUG] High memory consumption when running Claude Code in multiple IntelliJ projects
- [Bug] Auto-Compact Fails Due to Conversation Length Limit
- [BUG] OPUS TOKENS RUNNING WITHOUT OPUS BEING SELECTED
- [BUG] 'CLAUDE_CODE_MAX_OUTPUT_TOKENS' incorrectly validates Sonnet models against 32k limit instead of 64k
- Feature Request: Additional Claude Code Hook Events for Better Automation
- [BUG] Desktop app ignores `permissions.allow` rules — prompts for everything, even Claude's own config directory
- Remote Control: slash commands (/clear, /compact, /context, /bug) route to model as plain text instead of executing
- [BUG] v2.1.141: "Unhandled case: [object Object]" banner + permanent "Thinking…" stall in VS Code webview on image-bearing follow-up (Opus 4.7 1M)
- `--continue` on 2.1.90 silently drops conversation context (regression from 2.1.89)
- [MODEL] Claude uses Bash (cat, grep, head) instead of dedicated tools despite system prompt prohibition
- [BUG] "Clear context" option missing from plan mode exit dialog in --dangerously-skip-permissions mode
- Auto-compact not triggering on resumed sessions with Opus 4.6 (1M context)
- Opus 4.6 1M context window disappeared — was working, now gone
- [BUG] /resume crashes with "_.reduce is not a function" on large sessions (~2.7 MB, 1015 entries)
- [BUG] Anthropic Claude Model Degradation — Sourced Evidence
- Rules re-injected as system-reminders on every tool call, consuming ~46% of context window
- Add rate limit information to statusline data
- [Suspicious Behavior]: Hidden <system-reminder> 10,000+ injections consuming 15%+ of context window without user knowledge or consent
- [BUG] Two Claude Code sessions in the same folder / project share chat history when they should be separate.
- [BUG] [`Critical`] Compact issue block the normal use of Claude code
- Invalid Assistant Message Configuration During Compaction with Disabled Thinking
- Subagent model override silently dropped after a continuation boundary (SendMessage follow-up / post-compaction resume)
- Claude desktop local-agent VM (claudevm.bundle/rootfs.img) grows unboundedly and is never reclaimed — silently fills disk, causes out-of-space failures
- Agent output and permission-prompt rate increase as work becomes mechanical, inverse to cognitive load
- Unhandled case: [object Object]` banner on every turn in v2.1.141 (VS Code extension)
- [BUG] Single README.md edit burns entire 5-hour Max 5x window in 9m 39s (Opus 4.7 1M, resumed session, v2.1.126, JetBrains, Windows)
- CLI: Terminal scrollback permanently wiped during long tool-use chains
- Context compaction too aggressive on 1M context window (Opus 4.6)
- /clear creates a new session instead of clearing current context — misleading command semantics
- [BUG] API Error: 500 {"type":"error","error":{"type":"api_error","message":"Internal server error"},"request_id":"req_011CZ9DevVB5SScBuQrpfobu"}
- [BUG] Claude 1M Context Window — Advertised Capability Does Not Work as Marketed
- [FEATURE] Allow projects to exclude inherited .claude/rules/ from parent directories
- [Bug] Context limit reached without automatic compaction triggering
- [BUG] [1M] "Long context beta is not yet available for this subscription" error on Max plan ($200/mo)
- [BUG] context_window metrics not updated after /compact until next API interaction
- [BUG] Auto-Compact is not happening in Claude Code Cursor extension like it was before - goes past 100%
- [BUG] Error during compaction: Error: API Error: 400
- /compact fails, only auto-compact works.
- [Bug] Auto-Compact Triggering Excessively with Minimal Changes