Free Browser Tool

Claude Compaction Simulator

See how long Claude conversations can be summarized: what survives compaction, what becomes a short note, and what should be re-fetched or discarded.

Updated 2026-07-06 Free Client-side only All tools
Turn Kind Tokens Must survive

Goal, constraints, and definition of done should survive exactly.

 Decision

Keep provenance and exact facts; summarize surrounding prose.

 Evidence

Usually replace with pointer plus a short summary if it can be re-fetched.

 Tool result

Keep decisions, tradeoffs, paths, IDs, and unresolved risks.

 Decision

Usually drop or keep one sentence explaining why it was abandoned.

 Stale work

Discard unless it contains a durable user preference.

 Chatter

Raw history

0

After compaction

0

Recovered

0

Survives

    Lost or re-fetched

      How to use it

      1. Set the context window size and trigger threshold.
      2. Adjust the example turns or mark turns that must survive.
      3. Run the simulation to compare raw history with a compacted handoff.
      4. Use the survive/lost report to improve your own compaction prompt.

      Why this tool exists

      Claude context work often fails because teams reason from rough vibes instead of a visible budget. This tool makes the tradeoff explicit while keeping all inputs local to your browser.

      Treat the output as a planning estimate. For production routing, pair it with current Anthropic docs, real usage telemetry, and token counting against the exact model you call.

      FAQ

      Is this the actual Claude server-side compaction API?

      No. This is an educational simulator that models the tradeoff between raw history and summarized state.

      What should always survive compaction?

      Preserve user goals, hard constraints, decisions, exact IDs or paths, source provenance, and unresolved risks.

      What can usually be dropped?

      Drop small talk, duplicate tool output, outdated plans, and bulky results that can be fetched again from their source.

      Sources Used

      • Context windows, Claude Platform Docs. Accessed 2026-07-06. Official context-window behavior, token accumulation, extended thinking and tool-use accounting, and compaction guidance.
      • Context engineering: memory, compaction, and tool clearing, Claude Cookbook. Accessed 2026-07-06. Cookbook comparison of server-side compaction, context editing, tool-result clearing, and memory.
      • Effective context engineering for AI agents, Anthropic Engineering. Accessed 2026-07-06. Primary explanation of context engineering, context pollution, compaction, note-taking, and sub-agent architectures.