Agent-as-a-Graph: Knowledge Graph-Based Tool and Agent Retrieval for LLM Multi-Agent Systems

TL;DR

This paper introduces Agent-as-a-Graph, a knowledge graph-based retrieval method for multi-agent systems that improves agent and tool selection accuracy by representing and traversing agents and tools as a graph.

Contribution

It proposes a novel knowledge graph retrieval approach that enhances fine-grained agent and tool retrieval in large language model multi-agent systems.

Findings

Achieved 14.9% and 14.6% improvements in Recall@5 and nDCG@5 over previous methods.

Demonstrated effectiveness of type-specific weighted reciprocal rank fusion (wRRF).

Improved retrieval performance on the LiveMCPBenchmark.

Abstract

Recent advances in Large Language Model Multi-Agent Systems enable scalable orchestration and retrieval of specialized, parallelized subagents, each equipped with hundreds or thousands of Model Context Protocol (MCP) servers and tools. However, existing agent, MCP, and retrieval methods typically match queries against a single agent description, obscuring fine-grained tool capabilities of each agent, resulting in suboptimal agent selection. We introduce Agent-as-a-Graph retrieval, a knowledge graph retrieval augmented generation approach that represents both tools and their parent agents as nodes and edges in a knowledge graph. During retrieval, i) relevant agents and tool nodes are first retrieved through vector search, ii) we apply a type-specific weighted reciprocal rank fusion (wRRF) for reranking tools and agents, and iii) parent agents are traversed in the knowledge graph for the final set of agents. We evaluate Agent-as-a-Graph on the LiveMCPBenchmark, achieving 14.9% and 14.6% improvements in Recall@5 and nDCG@5 over prior state-of-the-art retrievers, and 2.4% improvements in wRRF optimizations.