MCPZoo Dataset Offers Largest Collection of Runnable Model Context Protocol Servers

Global: MCPZoo Dataset Offers Largest Collection of Runnable Model Context Protocol Servers

Researchers from several institutions announced on December 26, 2025 the public release of MCPZoo, a dataset that aggregates 129,059 Model Context Protocol (MCP) servers—56,053 of which are distinct—and includes 16,356 instances that have been deployed and verified as runnable. The effort aims to address the scarcity of large‑scale, accessible resources for studying MCP‑enabled AI agents.

Scope and Composition of the Dataset

The collection draws from multiple public repositories and code‑hosting platforms, consolidating over one hundred thousand server definitions. Each entry is accompanied by metadata describing its origin, configuration, and operational status. Of the total entries, 16,356 have undergone automated testing to confirm that they can be instantiated and interacted with in real time, providing a reliable foundation for experimental work.

Unified Access and Metadata Interfaces

MCPZoo supplies a standardized API and a set of metadata schemas that allow researchers to query, retrieve, and launch server instances without manual setup. The uniform interface abstracts away differences in underlying implementations, thereby reducing the overhead typically associated with deploying external tool servers for AI agents.

Advancing Research on AI Agent Tool Use

By offering a ready‑to‑use pool of MCP servers, the dataset enables systematic exploration of how autonomous agents can invoke external tools, a capability central to recent advances in tool‑augmented language models. Researchers can benchmark agent performance across a diverse set of services, ranging from data retrieval to specialized computation, without the need to develop individual integrations.

Facilitating Security and Vulnerability Analysis

The inclusion of verified, runnable servers also creates a practical testbed for cybersecurity investigations. Analysts can examine interaction patterns, assess isolation mechanisms, and probe for potential exploit vectors within MCP‑based architectures, contributing to a deeper understanding of associated risks.

Open Availability and Licensing

The MCPZoo dataset is released under an open‑access model with a DOI (10.48550/arXiv.2512.15144) and is hosted alongside the paper’s supplementary materials. Interested parties can download the full collection and accompanying documentation directly from the arXiv repository.

Future updates are planned to incorporate newly discovered servers and to refine verification procedures, encouraging community contributions and continuous improvement of the resource.

This report is based on information from arXiv, licensed under Academic Preprint / Open Access. Based on the abstract of the research paper. Full text available via ArXiv.