{"slug": "why-i-have-my-orchestrator-compact-at-100k-tokens", "title": "Why I have my orchestrator compact at 100K tokens", "summary": "A developer who runs Claude Opus 5.5 agents with a 1M-token context window reports compacting the orchestrator at 100K tokens, citing observed drift past roughly 200K tokens and warning that orchestrator drift propagates through subagent briefs. The writeup points to Anthropic's session-management guidance and Chroma's July 2025 context-rot evaluation of 18 LLMs, which found performance grows increasingly unreliable as input length grows, and notes that prompt caching cuts the price of 100K input tokens to as little as $0.02 on a cache hit without reducing the token count. It also flags compaction's downsides, including lossy summaries and broken prefix-cache reuse, and proposes a repeatable comparison logging input and cache token counts.", "body_md": "Claude Opus 5.5 [has a 1M token context window](https://platform.claude.com/docs/en/models/opus-5-5/overview). I keep every agent under 100K anyway. Past 200K a slight drift sets in.\n\nThese are personal observations, not measured thresholds.\n\nWhen a subagent drifts, you can get a weaker answer back. When the orchestrator drifts, that drift can spread through its briefs.\n\nClaude Code's subagents normally [start with a fresh, isolated context window](https://code.claude.com/docs/en/sub-agents). Forks inherit the parent conversation instead. A fresh subagent gets its task through the orchestrator's brief. Without the history, it may miss a constraint the brief dropped.\n\n[Anthropic's session guide](https://claude.com/resources/articles/using-claude-code-session-management-and-1m-context) warns that growing context can hurt performance. Claude Code summarizes automatically near the window's limit. The guide recommends compacting earlier and saying what comes next.\n\n```\n/context\n/compact focus on the auth refactor, drop the test debugging\n```\n\n[`/context`](https://code.claude.com/docs/en/commands) shows what's filling the window. The `/compact` line is the guide's own example. Codex has [`/compact`](https://learn.chatgpt.com/docs/developer-commands?surface=cli) too.\n\n[Chroma's Kelly Hong, Anton Troynikov and Jeff Huber](https://www.trychroma.com/research/context-rot) evaluated 18 LLMs in July 2025. Performance \"grows increasingly unreliable as input length grows\". The models included GPT-4.1 and Claude 4, not Opus 5.5. Neither source draws a line at 100K or 200K.\n\nFor 100K input tokens on Opus 5.5, using [API prices checked on November 3, 2026](https://platform.claude.com/docs/en/build-with-claude/prompt-caching):\n\n| Billed as | Per million tokens | 100K tokens | \n|---|---|---|\n| Uncached input | $4 | $0.40 | \n| Cache write, 5 minutes | $5 | $0.50 | \n| Cache write, 1 hour | $8 | $0.80 | \n| Cache hit | $0.20 | $0.02 | \n\nEach amount is the listed rate multiplied by one tenth. These are input costs only; output and tool charges are excluded. They are not subscription prices. Other models have other multipliers.\n\nThe hit is 0.05× the input price, and it doesn't buy room. [Anthropic's words](https://platform.claude.com/docs/en/build-with-claude/context-windows): \"prompt caching changes what you pay for those tokens, not whether they count.\"\n\nCompaction also rewrites the history. [Hits need 100 percent identical prompt segments](https://platform.claude.com/docs/en/build-with-claude/prompt-caching), so the unchanged start can still hit, and the rewritten part can't. [CompactionBench](https://www.microsoft.com/en-us/research/publication/benchmarking-long-horizon-agent-context-compaction-with-compactionbench/), from Microsoft researchers, lists \"disrupting prefix-cache reuse\" among the problems.\n\n[Amp removed compaction](https://ampcode.com/news/handoff) on October 23, 2025 and replaced it with Handoff. It calls compaction lossy: the agent decides what survives in the summary. Compacting early still carries that risk. A constraint you stated early can disappear.\n\nThis comparison has not been run for this article.\n\n`/context``/compact`. Restore the repo snapshot before each run. Repeat each comparison.\n\n```\ntask,model,effort,input_tokens,compacted,run,tests_passed,constraints_kept\n```\n\nOn the API, log [`input_tokens`, `cache_read_input_tokens` and `cache_creation_input_tokens`](https://platform.claude.com/docs/en/build-with-claude/prompt-caching). Those separate uncached input from cache reads and writes.\n\nThe maintained article, with all ten context terms and what each does to the budget, is [on argbe.tech](https://argbe.tech/blog/agent-context-compaction-100k?utm_source=devto&utm_medium=syndication&utm_campaign=agent-context-compaction-100k).", "url": "https://wpnews.pro/news/why-i-have-my-orchestrator-compact-at-100k-tokens", "canonical_source": "https://dev.to/argbetech/why-i-have-my-orchestrator-compact-at-100k-tokens-39mg", "published_at": "2026-11-16 20:12:00+00:00", "updated_at": "2026-11-16 20:21:15.704856+00:00", "lang": "en", "topics": ["ai-agents", "large-language-models", "ai-tools", "ai-infrastructure"], "entities": ["Anthropic", "Claude Opus 5.5", "Claude Code", "Chroma", "Kelly Hong", "Anton Troynikov", "Jeff Huber", "Amp"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/why-i-have-my-orchestrator-compact-at-100k-tokens", "markdown": "https://wpnews.pro/news/why-i-have-my-orchestrator-compact-at-100k-tokens.md", "text": "https://wpnews.pro/news/why-i-have-my-orchestrator-compact-at-100k-tokens.txt", "jsonld": "https://wpnews.pro/news/why-i-have-my-orchestrator-compact-at-100k-tokens.jsonld"}}