YT · YOUTUBE// 26d agoPRODUCT UPDATE

OpenAI API adds on-demand Tool Search

OpenAI introduced Tool Search in the API so agents can defer large tool definitions and load only the tools needed at runtime. This reduces up-front context bloat, helps token efficiency and cache preservation, and currently works on GPT-5.4+ models.

// ANALYSIS

This is a practical agent-engineering upgrade: it turns tool discovery into a runtime primitive instead of a prompt-tax developers pay every call.

–Tool Search supports deferred functions, namespaces, and MCP servers, so teams can keep tool catalogs large without always paying full schema cost.
–Hosted and client-executed modes cover both simple setups and dynamic tenant/project-specific tool loading patterns.
–OpenAI’s guidance to group tools into clear namespaces suggests retrieval quality matters as much as raw token savings.
–Official docs: https://developers.openai.com/api/docs/guides/tools-tool-search and SDK guidance: https://openai.github.io/openai-agents-python/tools/

// TAGS

openai-api-tool-searchopenai-apiagentapidevtoolmcpautomation

DISCOVERED

26d ago

2026-03-17

PUBLISHED

26d ago

2026-03-17

RELEVANCE

8/ 10

AUTHOR

Prompt Engineering