Intel / article

Harness Engineering in LLM Tool Use via Agent-Native Reusable Tool Primitives

Source: arxiv.org
Author: Haibo Jin, Suijin Wang, Xucheng Yu, Haojing Luo, Haohan Wang
Date: Published Sep 3, 2026

Context

HEART proposes making software tools easier for AI agents to combine by putting a language-model interface around each tool. The describes what it needs in natural language; the wrapper handles the tool's required input format and execution. The paper pairs these wrappers with ToolFace, a repository from which the agent retrieves relevant functions instead of loading every tool's interface into its working .

A coordinating system plans calls, routes them and checks results so it can recover from failures. The design moves some integration work into reusable tool wrappers, addressing both incompatible interfaces and the difficulty of choosing among large tool collections.

Terms in this piece · Glossary

AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.

Read the source arxiv.org