Content
# <img src="assets/icon.png" alt="MCP-Universe" width="23" height="23"> MCP-Universe
[](https://arxiv.org/abs/2508.14704)
[](https://mcp-universe.github.io/)
[](https://mcp-universe.github.io/#results)
[](https://discord.gg/t9tU77GF)
### 🎉 Latest Updates
> **📊 [MCPMark Evaluation](#mcpmark-benchmark)** - MCP-Universe now supports evaluating the MCPMark tasks
>
> **🚀 [MCP+](#mcp-precision-context-management-for-mcp-agents)** - Agentic wrapper on MCP clients which reduce token costs by up to 75%
>
> **🔬 [Deep Research Agent](#deep-research-agent-wide--deep-wd-research)** - Scale the Width of Deep Research Agents with parallel tool calling, improving performance and efficiency
</div>
---
## What is MCP-Universe?
MCP-Universe is a comprehensive ecosystem for building, optimizing, and evaluating AI agents that interact with the Model Context Protocol (MCP). Beyond our industry-leading benchmark for real-world MCP server interactions, MCP-Universe provides production-ready tools for agent development including specialized research agents ([**Deep Research Agent**](#deep-research-agent-wide--deep-wd-research)), intelligent context management ([**MCP+**](#mcp-precision-context-management-for-mcp-agents)), and sophisticated orchestration workflows.
<div align="center">
</div>
**Benchmarking:** Unlike existing benchmarks that rely on overly simplistic tasks, MCP-Universe addresses critical gaps by evaluating LLMs in **real-world scenarios** through interaction with actual MCP servers, capturing real application challenges such as:
- 🎯 **Long-horizon reasoning** across multi-step tasks
- 🔧 **Large, unfamiliar tool spaces** with diverse MCP servers
- 🌍 **Real-world data sources** and live environments
- ⚡ **Dynamic evaluation** with time-sensitive ground truth
## Table of Contents
- [What's New](#whats-new)
- [Architecture Overview](#architecture-overview)
- [Getting Started](#getting-started)
- [Prerequisites](#prerequisites)
- [Installation](#installation)
- [Quick Test](#quick-test)
- [Evaluating LLMs and Agents](#evaluating-llms-and-agents)
- [Prerequisites](#prerequisites-1)
- [Environment Configuration](#environment-configuration)
- [Benchmark Configuration](#benchmark-configuration)
- [Execution](#execution)
- [Save the running log](#save-the-running-log)
- [Save the benchmark result to a report](#save-the-benchmark-result-to-a-report)
- [Visualize the agent running information](#visualize-the-agent-running-information)
- [Creating Custom Benchmarks](#creating-custom-benchmarks)
- [Task definition](#task-definition)
- [Benchmark definition](#benchmark-definition)
- [Citation](#citation)
## What's New
### MCPMark Benchmark
**📊 Evaluate MCP Agents with MCPMark**
MCP-Universe now supports evaluating the **MCPMark** benchmark, enabling comprehensive testing and benchmarking of MCP agents. You can run MCPMark evaluations directly within the MCP-Universe framework to assess agent performance on MCP tasks.
**📚 Resources:**
- [How to run MCPMark](mcpuniverse/benchmark/configs/mcpmark/README.md#running-mcpmark-tasks)
- [Evaluation Scores](mcpuniverse/benchmark/configs/mcpmark/README.md#benchmark-results-alignment)
---
### MCP+: Precision Context Management for MCP Agents
**🚀 Reduce LLM Token Costs by up to 75% Without Sacrificing Quality**
MCP tools often return large, verbose outputs that waste your LLM's context window and cost money. **MCP+** wraps your MCP clients with intelligent post-processing that extracts only the relevant information before it reaches your LLM.
#### ✨ Key Features
- **💰 Massive Cost Reduction**: 50-75% token savings on tool outputs
- **⚡ Zero Code Changes**: Drop-in replacement for standard MCP clients
**📚 [Learn More at mcp-plus.github.io →](https://mcp-plus.github.io)**
</div>
---
### Deep Research Agent: Wide & Deep (W&D) Research
**🔬 Scale Research Width with Parallel Tool Calls**
**Feb 11, 2026** — We introduce **Wide & Deep (W&D) research agents** that scale *width* by making more parallel tool calls per turn. This approach improves accuracy on BrowseComp, HLE, and GAIA benchmarks while reducing turns, API cost, and wall-clock time. Our W&D agent with GPT-5-medium reaches **62.2%** on BrowseComp, outperforming GPT-5-high deep research (54.9%).
**📚 Resources:**
- [Paper](https://arxiv.org/pdf/2602.07359)
- [Website](https://xqlin98.github.io/wide-deep-research-agent/)
- [Code](mcpuniverse/benchmark/configs/deepresearch/README.md)
---
## Architecture Overview
The MCPUniverse architecture consists of the following key components:
- **Agents** (`mcpuniverse/agent/`): Base implementations for different agent types
- **Workflows** (`mcpuniverse/workflows/`): Orchestration and coordination layer
- **MCP Servers** (`mcpuniverse/mcp/`): Protocol management and external service integration
- **LLM Integration** (`mcpuniverse/llm/`): Multi-provider language model support
- **Benchmarking** (`mcpuniverse/benchmark/`): Evaluation and testing framework
- **Dashboard** (`mcpuniverse/dashboard/`): Visualization and monitoring interface
The diagram below illustrates the high-level view:
```
┌─────────────────────────────────────────────────────────────────┐
│ Application Layer │
├─────────────────────────────────────────────────────────────────┤
│ Dashboard │ Web API │ Python Lib │ Benchmarks │
│ (Gradio) │ (FastAPI) │ │ │
└─────────────┬─────────────────┬────────────────┬────────────────┘
│ │ │
┌─────────────▼─────────────────▼────────────────▼────────────────┐
│ Orchestration Layer │
├─────────────────────────────────────────────────────────────────┤
│ Workflows │ Benchmark Runner │
│ (Chain, Router, etc.) │ (Evaluation Engine) │
└─────────────┬─────────────────┬────────────────┬────────────────┘
│ │ │
┌─────────────▼─────────────────▼────────────────▼────────────────┐
│ Agent Layer │
├─────────────────────────────────────────────────────────────────┤
│ BasicAgent │ ReActAgent │ FunctionCall │ Other │
│ │ │ Agent │ Agents │
└─────────────┬─────────────────┬────────────────┬────────────────┘
│ │ │
┌─────────────▼─────────────────▼────────────────▼────────────────┐
│ Foundation Layer │
├─────────────────────────────────────────────────────────────────┤
│ MCP Manager │ LLM Manager │ Memory Systems │ Tracers │
│ (Servers & │ (Multi-Model │ (RAM, Redis) │ (Logging) │
│ Clients) │ Support) │ │ │
└─────────────────┴─────────────────┴─────────────────┴───────────┘
```
More information can be found [here](https://github.com/SalesforceAIResearch/MCP-Universe/blob/main/docs).
## Getting Started
We follow
the [feature branch workflow](https://www.atlassian.com/git/tutorials/comparing-workflows/feature-branch-workflow)
in this repo for its simplicity. To ensure code quality, [PyLint](https://pylint.readthedocs.io/en/latest/)
is integrated into our CI to enforce Python coding standards.
### Prerequisites
* **Python**: Requires version 3.10 or higher.
* **Docker**: Used for running Dockerized MCP servers.
* **PostgreSQL** (optional): Used for database storage and persistence.
* **Redis** (optional): Used for caching and memory management.
### Installation
1. **Clone the repository**
```bash
git clone https://github.com/SalesforceAIResearch/MCP-Universe.git
cd MCP-Universe
```
2. **Create and activate virtual environment**
```bash
python3 -m venv venv
source venv/bin/activate
```
3. **Install dependencies**
```bash
pip install -r requirements.txt
pip install -r dev-requirements.txt
```
4. **Platform-specific requirements**
**Linux:**
```bash
sudo apt-get install libpq-dev
```
**macOS:**
```bash
brew install postgresql
```
5. **Configure pre-commit hooks**
```bash
pre-commit install
```
6. **Environment configuration**
```bash
cp .env.example .env
# Edit .env with your API keys and configuration
```
### Quick Test
To run benchmarks, you first need to set environment variables:
1. Copy the `.env.example` file to a new file named `.env`.
2. In the `.env` file, set the required API keys for various services used by the agents,
such as `OPENAI_API_KEY` and `GOOGLE_MAPS_API_KEY`.
To execute a benchmark programmatically:
```python
from mcpuniverse.tracer.collectors import MemoryCollector # You can also use SQLiteCollector
from mcpuniverse.benchmark.runner import BenchmarkRunner
async def test():
trace_collector = MemoryCollector()
# Choose a benchmark config file under the folder "mcpuniverse/benchmark/configs"
benchmark = BenchmarkRunner("dummy/benchmark_1.yaml")
# Run the specified benchmark
results = await benchmark.run(trace_collector=trace_collector)
# Get traces
trace_id = results[0].task_trace_ids["dummy/tasks/weather_1.json"]
trace_records = trace_collector.get(trace_id)
```
## Evaluating LLMs and Agents
This section provides comprehensive instructions for evaluating LLMs and AI agents using the MCP-Universe benchmark suite. The framework supports evaluation across multiple domains including web search, location navigation, browser automation, financial analysis, repository management, and 3D design.
### Prerequisites
Before running benchmark evaluations, ensure you have completed the [Getting Started](#getting-started) section and have the following:
- Python: Version 3.10 or higher
- Docker: Installed and available in your environment
- All required dependencies installed via `pip install -r requirements.txt`
- Active virtual environment
- Appropriate API access for the services you intend to evaluate
### Environment Configuration
#### 1. Initial Setup
Copy the environment template and configure your API credentials:
```bash
cp .env.example .env
```
#### 2. API Keys and Configuration
Configure the following environment variables in your `.env` file. The required keys depend on which benchmark domains you plan to evaluate:
##### Core LLM Providers
| Environment Variable | Provider | Description | Required For |
|---------------------|----------|-------------|--------------|
| `OPENAI_API_KEY` | OpenAI | API key for GPT models (gpt-5, etc.) | All domains |
| `ANTHROPIC_API_KEY` | Anthropic | API key for Claude models | All domains |
| `GEMINI_API_KEY` | Google | API key for Gemini models | All domains |
> **Note**: You only need to configure the API key for the LLM provider you intend to use in your evaluation.
##### Domain-Specific Services
| Environment Variable | Service | Description | Setup Instructions |
|---------------------|---------|-------------|-------------------|
| `SERP_API_KEY` | SerpAPI | Web search API for search benchmark evaluation | [Get API key](https://serpapi.com/) |
| `GOOGLE_MAPS_API_KEY` | Google Maps | Geolocation and mapping services | [Setup Guide](https://console.cloud.google.com/google/maps-apis/credentials) |
| `GITHUB_PERSONAL_ACCESS_TOKEN` | GitHub | Personal access token for repository operations | [Token Setup](https://docs.github.com/en/authentication/keeping-your-account-and-data-secure/managing-your-personal-access-tokens) |
| `GITHUB_PERSONAL_ACCOUNT_NAME` | GitHub | Your GitHub username | N/A |
| `NOTION_API_KEY` | Notion | Integration token for Notion workspace access | [Integration Setup](https://developers.notion.com/docs/authorization#obtaining-a-token) |
| `NOTION_ROOT_PAGE` | Notion | Root page ID for your Notion workspace | See configuration example below |
##### System Paths
| Environment Variable | Description | Example |
|---------------------|-------------|---------|
| `BLENDER_APP_PATH` | Full path to Blender executable (we used v4.4.0) | `/Applications/Blender.app/Contents/MacOS/Blender` |
| `MCPUniverse_DIR` | Absolute path to your MCP-Universe repository | `/Users/username/MCP-Universe` |
##### Configuration Examples
**Notion Root Page ID:**
If your Notion page URL is:
```
https://www.notion.so/your_workspace/MCP-Evaluation-1dd6d96e12345678901234567eaf9eff
```
Set `NOTION_ROOT_PAGE=MCP-Evaluation-1dd6d96e12345678901234567eaf9eff`
**Blender Installation:**
1. Download Blender v4.4.0 from [blender.org](https://www.blender.org/)
2. Install our modified Blender MCP server following the [installation guide](docs/blender-setup.md)
3. Set the path to the Blender executable
##### ⚠️ Security Recommendations
> **🔒 IMPORTANT SECURITY NOTICE**
>
> Please read and follow these security guidelines carefully before running benchmarks:
- **🚨 GitHub Integration**: **CRITICAL** - We strongly recommend using a dedicated test GitHub account for benchmark evaluation. The AI agent will perform real operations on GitHub repositories, which could potentially modify or damage your personal repositories.
- **🔐 API Key Management**:
- Store API keys securely and never commit them to version control
- Use environment variables or secure key management systems
- Regularly rotate your API keys for enhanced security
- **🛡️ Access Permissions**:
- Grant minimal necessary permissions for each service integration
- Review and limit API key scopes to only required operations
- Monitor API usage and set appropriate rate limits
- **⚡ Blender Operations**: The 3D design benchmarks will execute Blender commands that may modify or create files on your system. Ensure you have adequate backups and run in an isolated environment if necessary.
### Benchmark Configuration
#### Domain-Specific Configuration Files
Each benchmark domain has a dedicated YAML configuration file located in `mcpuniverse/benchmark/configs/test/`. To evaluate your LLM/agent, modify the appropriate configuration file:
| Domain | Configuration File | Description |
|--------|-------------------|-------------|
| Web Search | `web_search.yaml` | Search engine and information retrieval tasks |
| Location Navigation | `location_navigation.yaml` | Geographic and mapping-related queries |
| Browser Automation | `browser_automation.yaml` | Web interaction and automation scenarios |
| Financial Analysis | `financial_analysis.yaml` | Market data analysis and financial computations |
| Repository Management | `repository_management.yaml` | Git operations and code repository tasks |
| 3D Design | `3d_design.yaml` | Blender-based 3D modeling and design tasks |
#### LLM Model Configuration
In each configuration file, update the LLM specification to match your target model:
```yaml
kind: llm
spec:
name: llm-1
type: openai # or anthropic, google, etc.
config:
model_name: gpt-4o # Replace with your target model
```
### Execution
#### Running Individual Benchmarks
Execute specific domain benchmarks using the following commands:
```bash
# Set Python path and run individual benchmarks
export PYTHONPATH=.
# Location Navigation
python tests/benchmark/mcpuniverse/test_benchmark_location_navigation.py
# Browser Automation
python tests/benchmark/mcpuniverse/test_benchmark_browser_automation.py
# Financial Analysis
python tests/benchmark/mcpuniverse/test_benchmark_financial_analysis.py
# Repository Management
python tests/benchmark/mcpuniverse/test_benchmark_repository_management.py
# Web Search
python tests/benchmark/mcpuniverse/test_benchmark_web_search.py
# 3D Design
python tests/benchmark/mcpuniverse/test_benchmark_3d_design.py
```
#### Batch Execution
For comprehensive evaluation across all domains:
```bash
#!/bin/bash
export PYTHONPATH=.
domains=("location_navigation" "browser_automation" "financial_analysis"
"repository_management" "web_search" "3d_design")
for domain in "${domains[@]}"; do
echo "Running benchmark: $domain"
python "tests/benchmark/mcpuniverse/test_benchmark_${domain}.py"
echo "Completed: $domain"
done
```
### Save the running log
If you want to save the running log, you can pass the `trace_collector` to the benchmark run function:
```python
from mcpuniverse.tracer.collectors import FileCollector
trace_collector = FileCollector(log_file="log/location_navigation.log")
benchmark_results = await benchmark.run(trace_collector=trace_collector)
```
### Save the benchmark result to a report
If you want to save a report of the benchmark result, you can use `BenchmarkReport` to dump a report:
```python
from mcpuniverse.benchmark.report import BenchmarkReport
report = BenchmarkReport(benchmark, trace_collector=trace_collector)
report.dump()
```
### Visualize the agent running information
To run the benchmark with intermediate results and see real-time progress, pass `callbacks=get_vprint_callbacks()` to the run function:
```python
from mcpuniverse.callbacks.handlers.vprint import get_vprint_callbacks
benchmark_results = await benchmark.run(
trace_collector=trace_collector,
callbacks=get_vprint_callbacks()
)
```
This will print out the intermediate results as the benchmark runs.
For further details, refer to the in-code documentation or existing configuration samples in the repository.
## Creating Custom Benchmarks
A benchmark is defined by three main configuration elements: the task definition,
agent/workflow definition, and the benchmark configuration itself. Below is an example
using a simple "weather forecasting" task.
### Task definition
The task definition is provided in JSON format, for example:
```json
{
"category": "general",
"question": "What's the weather in San Francisco now?",
"mcp_servers": [
{
"name": "weather"
}
],
"output_format": {
"city": "<City>",
"weather": "<Weather forecast results>"
},
"evaluators": [
{
"func": "json -> get(city)",
"op": "=",
"value": "San Francisco"
}
]
}
```
Field descriptions:
1. **category**: The task category, e.g., "general", "google-maps", etc. You can set any value for this property.
2. **question**: The main question you want to ask in this task. This is treated as a user message.
3. **mcp_servers**: A list of MCP servers that are supported in this framework.
4. **output_format**: The desired output format of agent responses.
5. **evaluators**: A list of tests to evaluate. For each test/evaluator, it has three attributes: "func" indicates
how to extract values from the agent response, "op" is the comparison operator, and "value" is the ground-truth
value.
It will evaluate **op(func(...), value, op_args...)**. "op" can be "=", "<", ">" or other customized operators.
In "evaluators", you need to write a rule ("func" attribute) showing how to extract values for testing. In the example
above, "json -> get(city)" will first do JSON decoding and then extract the value of key "city". There are several
predefined funcs in this repo:
1. **json**: Perform JSON decoding.
2. **get**: Get the value of a key.
3. **len**: Get the length of a list.
4. **foreach**: Do a FOR-EACH loop.
For example, let's define
```python
data = {"x": [{"y": [1]}, {"y": [1, 1]}, {"y": [1, 2, 3, 4]}]}
```
Then `get(x) -> foreach -> get(y) -> len` will do the following:
1. Get the value of "x": `[{"y": [1]}, {"y": [1, 1]}, {"y": [1, 2, 3, 4]}]`.
2. Do a foreach loop and get the value of "y": `[[1], [1, 1], [1, 2, 3, 4]]`.
3. Get the length of each list: `[1, 2, 4]`.
If these predefined functions are not enough, you can implement custom ones.
For more details, please check
this [doc](https://github.com/SalesforceAIResearch/MCP-Universe/blob/main/docs/custom-evaluators-guide.md).
### Benchmark definition
Define agent(s) and benchmark in a YAML file. Here’s a simple weather forecast benchmark:
```yaml
kind: llm
spec:
name: llm-1
type: openai
config:
model_name: gpt-4o
---
kind: agent
spec:
name: ReAct-agent
type: react
config:
llm: llm-1
instruction: You are an agent for weather forecasting.
servers:
- name: weather
---
kind: benchmark
spec:
description: Test the agent for weather forecasting
agent: ReAct-agent
tasks:
- dummy/tasks/weather.json
```
The benchmark definition mainly contains two parts: the agent definition and the benchmark configuration. The benchmark configuration is simple—you just need to specify the agent to use (by the defined agent name) and a list of tasks to evaluate. Each task entry is the task config file
path. It can be a full file path or a partial file path. If it is a partial file path (like "dummy/tasks/weather.json"),
it should be put in the
folder [mcpuniverse/benchmark/configs](https://github.com/SalesforceAIResearch/MCP-Universe/tree/main/mcpuniverse/benchmark/configs)
in this repo.
This framework offers a flexible way to define both simple agents (such as ReAct) and more complex, multi-step agent
workflows.
1. **Specify LLMs:** Begin by declaring the large language models (LLMs) you want the agents to use. Each LLM component
must be assigned a unique name (e.g., `"llm-1"`). These names serve as identifiers that the framework uses to connect
the different components together.
2. **Define an agent:** Next, define an agent by providing its name and selecting an agent class. Agent classes are
available in
the [mcpuniverse.agent](https://github.com/SalesforceAIResearch/MCP-Universe/tree/main/mcpuniverse/agent) package.
Commonly used classes include `"basic"`, `"function-call"`, and `"react"`. Within the agent specification (
`spec.config`), you must also indicate which LLM instance the agent should use by setting the `"llm"` field.
3. **Create complex workflows:** Beyond simple agents, the framework supports the definition of sophisticated,
orchestrated workflows where multiple agents interact or collaborate to solve more complex tasks.
For example:
```yaml
kind: llm
spec:
name: llm-1
type: openai
config:
model_name: gpt-4o
---
kind: agent
spec:
name: basic-agent
type: basic
config:
llm: llm-1
instruction: Return the latitude and the longitude of a place.
---
kind: agent
spec:
name: function-call-agent
type: function-call
config:
llm: llm-1
instruction: You are an agent for weather forecast. Please return the weather today at the given latitude and longitude.
servers:
- name: weather
---
kind: workflow
spec:
name: orchestrator-workflow
type: orchestrator
config:
llm: llm-1
agents:
- basic-agent
- function-call-agent
---
kind: benchmark
spec:
description: Test the agent for weather forecasting
agent: orchestrator-workflow
tasks:
- dummy/tasks/weather.json
```
## Citation
If you use MCP-Universe in your research, please cite our paper:
```bibtex
@misc{mcpuniverse,
title={MCP-Universe: Benchmarking Large Language Models with Real-World Model Context Protocol Servers},
author={Ziyang Luo and Zhiqi Shen and Wenzhuo Yang and Zirui Zhao and Prathyusha Jwalapuram and Amrita Saha and Doyen Sahoo and Silvio Savarese and Caiming Xiong and Junnan Li},
year={2025},
eprint={2508.14704},
archivePrefix={arXiv},
primaryClass={cs.AI},
url={https://arxiv.org/abs/2508.14704},
}
```
