Give your agent a memory.
Store agent context, continue conversations, and test recall in a fresh thread.
One assistant keeps its memories across conversations.
Assistant & thread
Start with empty IDs to let Backboard create a session, or resume an existing one.
Active assistant: not created yet · Active thread: new on next message. Only these IDs are saved in this browser; messages and memory live on Backboard.
A conversation with context
Memory mode is sent on every message.
Memory-only ingestion avoids a chat-model call. Use Auto to save context.
Off disables persistent memory, but existing thread history still provides context. Use a new thread to compare recall modes.
Your conversation
Provider results and retrieved context.
Context starts here
Connect your key and send a message, or select an assistant to inspect its memories.
⌘ Last request & responseView details +
LAST REQUEST
Run an action to inspect its request.
LAST RESPONSE
No response yet.
Backboard is a hosted service. Sending messages or storing context transmits that text to Backboard and may use provider credits. New thread keeps assistant memories; it does not erase remote conversations. Memory updates can be asynchronous—check their status before testing recall. This playground does not automatically attach memory to the other Gemma or Ollama pages.