> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.sparkian.com/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.sparkian.com/_mcp/server.

# Long context notice

As a conversation grows, Sparkian has to resend the entire chat history with every new message so the AI can keep track of the conversation. The **long context notice** is a banner that appears once a conversation gets long enough that this starts costing you noticeably more Sparks.

![Long context notice banner](https://cdn.sparkian.com/docs/long-context-notice-banner.png)

## Why longer chats cost more

Every reply requires the AI to read and process the full conversation so far, not just your latest message. The more messages accumulate, the more text has to be processed for each new reply, which increases the compute (and Spark) cost per message. See [Sparks (credit system)](/sparks) for how that cost is estimated.

## When the notice appears

Once an active conversation crosses **20,000 tokens**, Sparkian shows the banner to warn you before costs increase further.

## How to reduce the cost

* **Start a fresh chat** when you switch topics, instead of continuing an old thread.
* **Keep threads focused** - move to a new chat once a task is done rather than letting one conversation sprawl.
* **Lower your chat memory** - reduce how many previous messages are retained in [Chat Settings](/memory-and-context) (default is 20).
* **Enable Compact Context mode** - a Chat Settings option that summarizes older messages to preserve context while using fewer tokens.