WonderAgentPlay

← Blog

Last updated 2026-10-19

Muse Wonder Context Window and Active Memory

How WonderAgentPlay treats Muse Wonder context and why Meta Active Memory is not available in this browser playground.

People searching Muse Wonder context window and Muse Wonder Active Memory usually want two different things: how much text one request can hold, and whether the model remembers you across sessions. This playground only promises the first kind of *in-session* context, and even that is capped by our product limits — not by an unverified vendor maximum.

Context in this playground

  • The server clamps input size, message count, and max output tokens. The slider in the UI cannot raise those caps.
  • A frontend token estimate is display-only. It does not decide billing or acceptance.
  • Long pasted files can be truncated or rejected. Split a task: reproduce the bug, then ask for a small diff, then ask for tests.

We do not publish a measured context-window number for meta/muse-wonder-1.0 from a successful call on this site. Dashboard marketing figures, if you see them elsewhere, are not a guarantee for this deployment.

Active Memory

Meta native Active Memory is not available here.

  • Conversation history is stored in this browser (localStorage) so a refresh can restore the thread.
  • Clearing site data, using another device, or another browser starts a new thread.
  • We do not sync chats to a Meta account, and we do not claim persistent agent memory.

If a future Muse product offers account-level memory, that would be a separate feature. It is not part of this phase.

Practical prompt shape

text
Context (only the relevant function):
...paste class="tok-number">40–class="tok-number">80 lines...

Task: name the bug, then propose a minimal patch.
Do not assume files you have not been shown.

Related: how to use Muse Wonder online and the JSON output example.

Try the related flow on the playground.

Muse Wonder Context Window and Active Memory · WonderAgentPlay