Theo-brainUnofficial. Not affiliated with Theo or t3.gg.
Sign inYou

T3 Code guide

Claude conversation compaction

Models and providers

Claude provider settings include Auto-compact after, which takes 100000 to 1000000 tokens. A separate setting controls how idle threads over 200,000 tokens are compacted.

T3 Code docs: providers claude

What it’s for

Summarizing large history when you return to a thread after Claude's prompt cache has expired, instead of resending the full history.

T3 Code docs: providers claude

How to use it

  1. Set Auto-compact after in the Claude provider settings, for example 300000. Leave it empty to use Claude Code's default.

    T3 Code docs: providers claude

  2. When the Compact chip shows next to Send, press Enter to summarize first, or click the chip to switch it to Keep.

    T3 Code docs: providers claude

  3. Under Compact idle threads, choose When you return, Before the cache expires (beta), or Never.

    T3 Code docs: providers claude

  4. Send /compact in a conversation to compact it by hand.

    T3 Code docs: providers claude

Theo’s tips

Don't worry about context size. The catch is an expired cache: if a large context isn't cached anymore, use 'resume with less context' and compact quickly.

37:46The one catch being if the cash has expired, which thankfully in tools like T3 chat, we now expose. They have this little resume with less context. If the 527,000 tokens from earlier aren't cached yet, might be worth compacting quick.

So I was using Fable wrong...

What’s changed

  • Idle compaction now has a global default with per-instance overrides.Updated Oct 21
  • Old Claude threads are compacted when you send, instead of stacking notices.Updated Oct 15
  • Old Claude threads compact on send instead of stacking notices.Updated Oct 7
1 earlier change
  • Old threads are compacted before they burn through usage.Added Aug 25

Related settings

Unofficial notes. Not affiliated with Theo or t3.gg.