Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ToolACE-MCP: Generalizing History-Aware Routing from MCP Tools to the Agent Web

Zhiyuan Yao, Zishan Xu, Yifu Guo, Zhiguang Han, Cheng Yang, Shuo Zhang, Weinan Zhang, Xingshan Zeng, Weiwen Liu

TL;DR

ToolACE-MCP introduces a history-aware router training framework to tackle the scalability and generality bottlenecks of tool-centric agent ecosystems. It builds a self-evolving candidate graph, synthesizes multi-turn trajectories via a multi-agent simulator, and trains a lightweight router that generalizes from tool routing to agent routing. The Light Routing Agent enables plug-and-play deployment, achieving strong results on MCP benchmarks and robust cross-domain generalization, including to multi-agent orchestration tasks. The work demonstrates significant improvements over embedding-based retrieval and ReAct baselines, and shows resilience to large candidate spaces and tool noise, establishing a practical foundation for universal orchestration in the Agent Web.

Abstract

With the rise of the Agent Web and Model Context Protocol (MCP), the agent ecosystem is evolving into an open collaborative network, exponentially increasing accessible tools. However, current architectures face severe scalability and generality bottlenecks. To address this, we propose ToolACE-MCP, a pipeline for training history-aware routers to empower precise navigation in large-scale ecosystems. By leveraging a dependency-rich candidate Graph to synthesize multi-turn trajectories, we effectively train routers with dynamic context understanding to create the plug-and-play Light Routing Agent. Experiments on the real-world benchmarks MCP-Universe and MCP-Mark demonstrate superior performance. Notably, ToolACE-MCP exhibits critical properties for the future Agent Web: it not only generalizes to multi-agent collaboration with minimal adaptation but also maintains exceptional robustness against noise and scales effectively to massive candidate spaces. These findings provide a strong empirical foundation for universal orchestration in open-ended ecosystems.

ToolACE-MCP: Generalizing History-Aware Routing from MCP Tools to the Agent Web

TL;DR

ToolACE-MCP introduces a history-aware router training framework to tackle the scalability and generality bottlenecks of tool-centric agent ecosystems. It builds a self-evolving candidate graph, synthesizes multi-turn trajectories via a multi-agent simulator, and trains a lightweight router that generalizes from tool routing to agent routing. The Light Routing Agent enables plug-and-play deployment, achieving strong results on MCP benchmarks and robust cross-domain generalization, including to multi-agent orchestration tasks. The work demonstrates significant improvements over embedding-based retrieval and ReAct baselines, and shows resilience to large candidate spaces and tool noise, establishing a practical foundation for universal orchestration in the Agent Web.

Abstract

With the rise of the Agent Web and Model Context Protocol (MCP), the agent ecosystem is evolving into an open collaborative network, exponentially increasing accessible tools. However, current architectures face severe scalability and generality bottlenecks. To address this, we propose ToolACE-MCP, a pipeline for training history-aware routers to empower precise navigation in large-scale ecosystems. By leveraging a dependency-rich candidate Graph to synthesize multi-turn trajectories, we effectively train routers with dynamic context understanding to create the plug-and-play Light Routing Agent. Experiments on the real-world benchmarks MCP-Universe and MCP-Mark demonstrate superior performance. Notably, ToolACE-MCP exhibits critical properties for the future Agent Web: it not only generalizes to multi-agent collaboration with minimal adaptation but also maintains exceptional robustness against noise and scales effectively to massive candidate spaces. These findings provide a strong empirical foundation for universal orchestration in open-ended ecosystems.
Paper Structure (29 sections, 5 equations, 4 figures, 4 tables)

This paper contains 29 sections, 5 equations, 4 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Large-Scale Tool Learning
  4. Dynamic Tool Routing
  5. Method
  6. Overview
  7. Problem Formulation.
  8. Candidate Graph-based Extension With Self-Evolutionary Mutation
  9. Graph Construction
  10. Self-Evolutionary Mutation
  11. History-Aware Supervision for Router Training
  12. Light Routing Agent
  13. Experiment
  14. Experiment Setup
  15. Dataset and model
  16. ...and 14 more sections

Figures (4)

  • Figure 1: Comparison of ToolACE-MCP with other existing paradigms. (a) Static Injection: Constrained by finite context windows and rigid schemas. (b) Embedding-based Retrieval: Limited by static semantic matching and lack of historical context awareness. (c) ToolACE-MCP (Ours): A robust router that leverages reasoning and interaction history to achieve high-accuracy retrieval within a massive candidate space.
  • Figure 2: The overall framework of ToolACE-MCP. It consists of three key stages: (1) Self-evolutionary Graph Construction, which expands and structures the candidate space via mutation and relation modeling; (2) Multi-Agent Simulation, which synthesizes interaction trajectories to extract history-aware supervision signals; and (3) The Light Routing Agent, designed to seamlessly integrate the trained router into the inference pipeline.
  • Figure 3: Performance evaluation on the Agent Route Benchmark. Comparative analysis of agent route accuracy between ToolACE-MCP and representative baselines.
  • Figure 4: Impact of Historical Context. A performance comparison between the history-aware model and an ablation variant trained without historical context.