MCP-Cosmos: World Model-Augmented Agents for Complex Task Execution in MCP Environments

Hugging Face Daily Papers Papers

Summary

This paper introduces MCP-Cosmos, a framework that integrates generative world models into the Model Context Protocol ecosystem to enhance agent planning and execution through predictive simulation in latent space.

The Model Context Protocol (MCP) has unified the interface between Large Language Models (LLMs) and external tools, yet a fundamental gap remains in how agents conceptualize the environments within which they operate. Current paradigms are bifurcated: Task-level planning often ignores execution-time dynamics, while reactive execution lacks long-horizon foresight. We present MCP-Cosmos, a framework that infuses generative World Models (WM) into the MCP ecosystem to enable predictive task automation. By unifying three disparate technologies, namely MCP, World Model, and Agent, we demonstrate that a "Bring Your Own World Model" (BYOWM) strategy allows agents to simulate state transitions and refine plans in a latent space before execution. We conducted experiments using two strategies, namely ReAct and SPIRAL with 2 planning models and 3 representative world models over 20+ MCP-Bench tasks. We observed improvements in Agent's environment interaction KPI such as tool success rate and tool parameter accuracy. The framework also offers new metrics such as Execution Quality to generate new insights about the effectiveness of world models compared to baseline.
Original Article
View Cached Full Text

Cached at: 05/13/26, 04:10 AM

Paper page - MCP-Cosmos: World Model-Augmented Agents for Complex Task Execution in MCP Environments

Source: https://huggingface.co/papers/2605.09131

Abstract

MCP-Cosmos integrates generative World Models into the Model Context Protocol ecosystem to enhance agent planning and execution through predictive simulation in latent space.

TheModel Context Protocol(MCP) has unified the interface betweenLarge Language Models(LLMs) and external tools, yet a fundamental gap remains in how agents conceptualize the environments within which they operate. Current paradigms are bifurcated: Task-level planning often ignores execution-time dynamics, whilereactive execution lacks long-horizon foresight. We present MCP-Cosmos, a framework that infusesgenerative World Models(WM) into the MCP ecosystem to enable predictivetask automation. By unifying three disparate technologies, namely MCP, World Model, and Agent, we demonstrate that a “Bring Your Own World Model” (BYOWM) strategy allows agents to simulate state transitions and refine plans in alatent spacebefore execution. We conducted experiments using two strategies, namelyReActandSPIRALwith 2 planning models and 3 representative world models over 20+ MCP-Bench tasks. We observed improvements in Agent’s environment interaction KPI such astool success rateandtool parameter accuracy. The framework also offers new metrics such asExecution Qualityto generate new insights about the effectiveness of world models compared to baseline.

View arXiv pageView PDFAdd to collection

Get this paper in your agent:

hf papers read 2605\.09131

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2605.09131 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2605.09131 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2605.09131 in a Space README.md to link it from this page.

Collections including this paper1

Similar Articles

Code execution with MCP: Building more efficient agents

Anthropic Engineering

This article from Anthropic explores how integrating code execution with the Model Context Protocol (MCP) can improve the efficiency of AI agents. It addresses challenges like token overload from tool definitions and intermediate results, proposing code execution as a solution to reduce latency and costs.

@swyx: explain this

X AI KOLs Following

This article explains the Model Context Protocol (MCP) for building pluggable AI agent architectures, detailing lessons from building an MCP server at Sentry, including OAuth 2.1 integration, designing agent-friendly tool interfaces, and current ecosystem limitations.

Contextberg

Product Hunt

Contextberg turns your work into AI agent memory, served over the Model Context Protocol (MCP).