Context management¶

Every conversation has a context window — the amount of text the model can consider at once, including your messages, the agent’s replies, tool results, and any instructions or files in scope. CoCo Desktop gives you visibility into how that window is being used and tools to keep it clear so long conversations stay responsive.

See the token breakdown¶

The context ring in the bottom-right of the composer (the message input) fills up as the conversation consumes the context window, so you can see at a glance how much room is left. Hover the ring to open the Context window popover, which shows the tokens used out of the total, the percentage used, and an estimated breakdown of what’s taking up space.

Compact a conversation¶

When a conversation grows long, run /compact from the chat input to summarize the history and free up context. Compacting replaces the earlier turns the model carries with a concise summary, so the agent keeps the thread of what you were doing without re-reading everything — and the full conversation stays visible to you on screen.

Use /compact when a session is getting long but you want to keep going in the same conversation rather than starting a new one.

Automatic token savings¶

CoCo also reduces token use on its own — for example, by loading your most-used tools up front and discovering the rest on demand, offloading large tool results to disk instead of your context, and sending only the new part of the conversation each turn. These behaviors are automatic and require no configuration.