Researchers introduced MCPAgentBench, a benchmark built from real-world tasks to measure how well LLM agents use tools through the Model Context Protocol. The authors note that existing MCP evaluation suites have limitations, which their benchmark aims to address. It targets assessment of practical tool-calling ability in autonomous agent settings.
The Model Context Protocol has been revised with a 2026-07-28 spec that replaces stateful sessions with a stateless communication model. According to the report, this is the most substantial change to the protocol since it was first published, and developers building new applications on it will need to rethink their existing design assumptions.
An arXiv paper argues that research into the Model Context Protocol server ecosystem relies on sampling methods that implicitly favor functioning servers, such as reference sets, popularity rankings, hand-picked collections, or pipelines that fix servers until they run. The authors examine what a random draw from the MCP registry actually returns and contrast it with the servers used in tool-use benchmarks. The work highlights how these sampling choices can distort findings about the ecosystem and about agent tool-use performance.
A new arXiv preprint describes an approach for finding indirect prompt injection vulnerabilities in MCP servers without needing system access or live interaction. The method works from descriptions alone, targeting third-party audits of closed-source, remotely hosted, or commercially gated software where conventional analysis is not possible. The authors position it as a way to assess critical systems that analysts cannot directly inspect or probe.