🔥 Feature Spotlight: Prompt caching in Make AI Agents

News
, ,
1

Prompt caching helps Make AI Agents avoid reprocessing stable instructions and tool definitions on repeated calls, so recurring agent work responds faster and uses fewer input tokens.

AI Agents often make several calls while working through a task. When stable instructions and tool definitions stay the same, prompt caching lets Make reuse them instead of processing them from scratch each time.

:brain: Prompt caching is available in Make AI Agents. For Anthropic Claude models, choose a 5-minute or 1-hour cache for your agent’s stable instructions and tool definitions. With OpenAI models, it works automatically.

:new_button: What’s new in Prompt Caching in Make AI Agents

  • Anthropic Claude models prompt caching: Choose the cache duration that fits your workflow (Off, 5 minutes, or 1 hour).
  • OpenAI models prompt caching: Caching is automatic. Make reuses the prompt content that can be cached, with nothing to configure.
  • Faster repeat calls: When an agent uses the same stable instructions and tools again, it can spend less time reprocessing them.
  • Lower input costs: Caching can reduce the input-token cost of repeat calls.

:gear: How to use Prompt caching in Make AI Agents

In the Make AI Agent module, click Show advanced settings. Prompt caching works automatically with OpenAI models. For Anthropic Claude models, choose a 5-minute or 1-hour cache to reuse your agent’s stable instructions and tool definitions.

2048×1285 278 KB

:flexed_biceps: Prompt caching in Make AI Agents use cases

  • Customer support workflows: Classify or route follow-up requests using the same agent setup.
  • Research and analysis workflows: Use the same instructions and tools across repeat agent calls.
  • Operations assistants: Handle recurring triage, checks, or routing with a consistent agent setup.
  • Knowledge-base assistants: reuse the same instructions and tool definitions while answering related employee or customer questions.
  • Lead and incident triage: classify and route incoming leads or incidents with a consistent agent setup.

Caching works best when the stable parts of your agent setup stay the same across calls.

Happy automating! :purple_circle:

Valery from Make

:books: Helpful resources

:backhand_index_pointing_right: Create your first AI agent
:backhand_index_pointing_right: Claude developer docs: Prompt caching
:backhand_index_pointing_right: OpenAI developer docs: Prompt caching

🔥 App update: Anthropic Claude models are available in Make AI Provider

4 Likes