Prompt Caching and LLM Cost

Prompt caching lets providers reuse expensive prefix computation across requests. This lowers latency and cost, especially for agents that run many tool calls in one conversation.

Cheap and fast patterns

• Keep instructions and skill definitions stable across calls.
• Load project context once, then reuse.
• Use summaries for long histories instead of full transcript resends.

GreatChat is designed around this workflow: assistants, memory, and connections keep context clean without losing capability.