mcp-use 1.2.13__tar.gz → 1.3.0__tar.gz

{mcp_use-1.2.13 → mcp_use-1.3.0}/.github/workflows/unittests.yml

@@ -22,7 +22,7 @@ jobs:
       - name: Install dependencies
         run: |
           python -m pip install --upgrade pip
-          pip install .[dev,anthropic,openai,search]
+          pip install .[dev,anthropic,openai,search,e2b]
       - name: Lint with ruff
         run: |
           ruff check .

{mcp_use-1.2.13 → mcp_use-1.3.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: mcp-use
-Version: 1.2.13
+Version: 1.3.0
 Summary: MCP Library for LLMs
 Author-email: Pietro Zullo <pietro.zullo@gmail.com>
 License: MIT
@@ -33,6 +33,8 @@ Requires-Dist: pytest-asyncio>=0.21.0; extra == 'dev'
 Requires-Dist: pytest-cov>=4.1.0; extra == 'dev'
 Requires-Dist: pytest>=7.4.0; extra == 'dev'
 Requires-Dist: ruff>=0.1.0; extra == 'dev'
+Provides-Extra: e2b
+Requires-Dist: e2b-code-interpreter>=1.5.0; extra == 'e2b'
 Provides-Extra: openai
 Requires-Dist: openai>=1.10.0; extra == 'openai'
 Provides-Extra: search
@@ -67,30 +69,64 @@ Description-Content-Type: text/markdown
         <img src="https://img.shields.io/github/stars/pietrozullo/mcp-use?style=social" /></a>
     </p>
     <p align="center">
-    <a href="https://x.com/pietrozullo" alt="Twitter Follow">
+    <a href="https://x.com/pietrozullo" alt="Twitter Follow - Pietro">
         <img src="https://img.shields.io/twitter/follow/Pietro?style=social" /></a>
+    <a href="https://x.com/pederzh" alt="Twitter Follow - Luigi">
+        <img src="https://img.shields.io/twitter/follow/Luigi?style=social" /></a>
     <a href="https://discord.gg/XkNkSkMz3V" alt="Discord">
         <img src="https://dcbadge.limes.pink/api/server/https://discord.gg/XkNkSkMz3V?style=flat" /></a>
 </p>
 🌐 MCP-Use is the open source way to connect **any LLM to any MCP server** and build custom agents that have tool access, without using closed source or application clients.
 
+💬 Get started quickly - chat with your servers on our <b>hosted version</b>! <b>[Try mcp-use chat *(beta)* ](https://chat.mcp-use.io)</b>.
+
 💡 Let developers easily connect any LLM to tools like web browsing, file operations, and more.
 
 # Features
 
 ## ✨ Key Features
-
-| Feature | Description |
-|---------|-------------|
-| 🔄 [**Ease of use**](#quick-start) | Create your first MCP capable agent you need only 6 lines of code |
-| 🤖 [**LLM Flexibility**](#installing-langchain-providers) | Works with any langchain supported LLM that supports tool calling (OpenAI, Anthropic, Groq, LLama etc.) |
-| 🌐 [**Code Builder**](https://mcp-use.io/builder) | Explore MCP capabilities and generate starter code with the interactive [code builder](https://mcp-use.io/builder). |
-| 🔗 [**HTTP Support**](#http-connection-example) | Direct connection to MCP servers running on specific HTTP ports |
-| ⚙️ [**Dynamic Server Selection**](#dynamic-server-selection-server-manager) | Agents can dynamically choose the most appropriate MCP server for a given task from the available pool |
-| 🧩 [**Multi-Server Support**](#multi-server-support) | Use multiple MCP servers simultaneously in a single agent |
-| 🛡️ [**Tool Restrictions**](#tool-access-control) | Restrict potentially dangerous tools like file system or network access |
-| 🔧 [**Custom Agents**](#build-a-custom-agent) | Build your own agents with any framework using the LangChain adapter or create new adapters |
-
+<table>
+  <tr>
+    <th width="400">Feature</th>
+    <th>Description</th>
+  </tr>
+  <tr>
+    <td>🔄 <a href="#quick-start"><strong>Ease of use</strong></a></td>
+    <td>Create your first MCP capable agent you need only 6 lines of code</td>
+  </tr>
+  <tr>
+    <td>🤖 <a href="#installing-langchain-providers"><strong>LLM Flexibility</strong></a></td>
+    <td>Works with any langchain supported LLM that supports tool calling (OpenAI, Anthropic, Groq, LLama etc.)</td>
+  </tr>
+  <tr>
+    <td>🌐 <a href="https://mcp-use.io/builder"><strong>Code Builder</strong></a></td>
+    <td>Explore MCP capabilities and generate starter code with the interactive <a href="https://mcp-use.io/builder">code builder</a>.</td>
+  </tr>
+  <tr>
+    <td>🔗 <a href="#http-connection-example"><strong>HTTP Support</strong></a></td>
+    <td>Direct connection to MCP servers running on specific HTTP ports</td>
+  </tr>
+  <tr>
+    <td>⚙️ <a href="#dynamic-server-selection-server-manager"><strong>Dynamic Server Selection</strong></a></td>
+    <td>Agents can dynamically choose the most appropriate MCP server for a given task from the available pool</td>
+  </tr>
+  <tr>
+    <td>🧩 <a href="#multi-server-support"><strong>Multi-Server Support</strong></a></td>
+    <td>Use multiple MCP servers simultaneously in a single agent</td>
+  </tr>
+  <tr>
+    <td>🛡️ <a href="#tool-access-control"><strong>Tool Restrictions</strong></a></td>
+    <td>Restrict potentially dangerous tools like file system or network access</td>
+  </tr>
+  <tr>
+    <td>🔧 <a href="#build-a-custom-agent"><strong>Custom Agents</strong></a></td>
+    <td>Build your own agents with any framework using the LangChain adapter or create new adapters</td>
+  </tr>
+  <tr>
+    <td>❓ <a href="https://mcp-use.io/what-should-we-build-next"><strong>What should we build next</strong></a></td>
+    <td>Let us know what you'd like us to build next</td>
+  </tr>
+</table>
 
 # Quick start
 
@@ -118,11 +154,8 @@ pip install langchain-openai
 
 # For Anthropic
 pip install langchain-anthropic
-
-# For other providers, check the [LangChain chat models documentation](https://python.langchain.com/docs/integrations/chat/)
 ```
-
-and add your API keys for the provider you want to use to your `.env` file.
+For other providers, check the [LangChain chat models documentation](https://python.langchain.com/docs/integrations/chat/) and add your API keys for the provider you want to use to your `.env` file.
 
 ```bash
 OPENAI_API_KEY=
@@ -561,6 +594,101 @@ if __name__ == "__main__":
     asyncio.run(main())
 ```
 
+# Sandboxed Execution
+
+MCP-Use supports running MCP servers in a sandboxed environment using E2B's cloud infrastructure. This allows you to run MCP servers without having to install dependencies locally, making it easier to use tools that might have complex setups or system requirements.
+
+## Installation
+
+To use sandboxed execution, you need to install the E2B dependency:
+
+```bash
+# Install mcp-use with E2B support
+pip install "mcp-use[e2b]"
+
+# Or install the dependency directly
+pip install e2b-code-interpreter
+```
+
+You'll also need an E2B API key. You can sign up at [e2b.dev](https://e2b.dev) to get your API key.
+
+## Configuration
+
+To enable sandboxed execution, use the `ClientOptions` parameter when creating your `MCPClient`:
+
+```python
+import asyncio
+import os
+from dotenv import load_dotenv
+from langchain_openai import ChatOpenAI
+from mcp_use import MCPAgent, MCPClient
+from mcp_use.types.sandbox import SandboxOptions
+from mcp_use.types.clientoptions import ClientOptions
+
+async def main():
+    # Load environment variables (needs E2B_API_KEY)
+    load_dotenv()
+
+    # Define MCP server configuration
+    server_config = {
+        "mcpServers": {
+            "everything": {
+                "command": "npx",
+                "args": ["-y", "@modelcontextprotocol/server-everything"],
+            }
+        }
+    }
+
+    # Define sandbox options
+    sandbox_options: SandboxOptions = {
+        "api_key": os.getenv("E2B_API_KEY"),  # API key can also be provided directly
+        "sandbox_template_id": "base",  # Use base template
+    }
+
+    # Create client options for sandboxed mode
+    client_options: ClientOptions = {
+        "is_sandboxed": True,
+        "sandbox_options": sandbox_options
+    }
+
+    # Create client with sandboxed mode enabled
+    client = MCPClient(
+        config=server_config,
+        options=client_options
+    )
+
+    # Create agent with the sandboxed client
+    llm = ChatOpenAI(model="gpt-4o")
+    agent = MCPAgent(llm=llm, client=client)
+
+    # Run your agent
+    result = await agent.run("Use the command line tools to help me add 1+1")
+    print(result)
+
+    # Clean up
+    await client.close_all_sessions()
+
+if __name__ == "__main__":
+    asyncio.run(main())
+```
+
+## Sandbox Options
+
+The `SandboxOptions` type provides configuration for the sandbox environment:
+
+| Option                 | Description                                                                              | Default               |
+| ---------------------- | ---------------------------------------------------------------------------------------- | --------------------- |
+| `api_key`              | E2B API key. Required - can be provided directly or via E2B_API_KEY environment variable | None                  |
+| `sandbox_template_id`  | Template ID for the sandbox environment                                                  | "base"                |
+| `supergateway_command` | Command to run supergateway                                                              | "npx -y supergateway" |
+
+## Benefits of Sandboxed Execution
+
+- **No local dependencies**: Run MCP servers without installing dependencies locally
+- **Isolation**: Execute code in a secure, isolated environment
+- **Consistent environment**: Ensure consistent behavior across different systems
+- **Resource efficiency**: Offload resource-intensive tasks to cloud infrastructure
+
 # Build a Custom Agent:
 
 You can also build your own custom agent using the LangChain adapter:
@@ -652,7 +780,6 @@ agent = MCPAgent(
 
 This is useful when you only need to see the agent's steps and decision-making process without all the low-level debug information from other components.
 
-
 # Roadmap
 
 <ul>

{mcp_use-1.2.13 → mcp_use-1.3.0}/README.md

@@ -26,30 +26,64 @@
         <img src="https://img.shields.io/github/stars/pietrozullo/mcp-use?style=social" /></a>
     </p>
     <p align="center">
-    <a href="https://x.com/pietrozullo" alt="Twitter Follow">
+    <a href="https://x.com/pietrozullo" alt="Twitter Follow - Pietro">
         <img src="https://img.shields.io/twitter/follow/Pietro?style=social" /></a>
+    <a href="https://x.com/pederzh" alt="Twitter Follow - Luigi">
+        <img src="https://img.shields.io/twitter/follow/Luigi?style=social" /></a>
     <a href="https://discord.gg/XkNkSkMz3V" alt="Discord">
         <img src="https://dcbadge.limes.pink/api/server/https://discord.gg/XkNkSkMz3V?style=flat" /></a>
 </p>
 🌐 MCP-Use is the open source way to connect **any LLM to any MCP server** and build custom agents that have tool access, without using closed source or application clients.
 
+💬 Get started quickly - chat with your servers on our <b>hosted version</b>! <b>[Try mcp-use chat *(beta)* ](https://chat.mcp-use.io)</b>.
+
 💡 Let developers easily connect any LLM to tools like web browsing, file operations, and more.
 
 # Features
 
 ## ✨ Key Features
-
-| Feature | Description |
-|---------|-------------|
-| 🔄 [**Ease of use**](#quick-start) | Create your first MCP capable agent you need only 6 lines of code |
-| 🤖 [**LLM Flexibility**](#installing-langchain-providers) | Works with any langchain supported LLM that supports tool calling (OpenAI, Anthropic, Groq, LLama etc.) |
-| 🌐 [**Code Builder**](https://mcp-use.io/builder) | Explore MCP capabilities and generate starter code with the interactive [code builder](https://mcp-use.io/builder). |
-| 🔗 [**HTTP Support**](#http-connection-example) | Direct connection to MCP servers running on specific HTTP ports |
-| ⚙️ [**Dynamic Server Selection**](#dynamic-server-selection-server-manager) | Agents can dynamically choose the most appropriate MCP server for a given task from the available pool |
-| 🧩 [**Multi-Server Support**](#multi-server-support) | Use multiple MCP servers simultaneously in a single agent |
-| 🛡️ [**Tool Restrictions**](#tool-access-control) | Restrict potentially dangerous tools like file system or network access |
-| 🔧 [**Custom Agents**](#build-a-custom-agent) | Build your own agents with any framework using the LangChain adapter or create new adapters |
-
+<table>
+  <tr>
+    <th width="400">Feature</th>
+    <th>Description</th>
+  </tr>
+  <tr>
+    <td>🔄 <a href="#quick-start"><strong>Ease of use</strong></a></td>
+    <td>Create your first MCP capable agent you need only 6 lines of code</td>
+  </tr>
+  <tr>
+    <td>🤖 <a href="#installing-langchain-providers"><strong>LLM Flexibility</strong></a></td>
+    <td>Works with any langchain supported LLM that supports tool calling (OpenAI, Anthropic, Groq, LLama etc.)</td>
+  </tr>
+  <tr>
+    <td>🌐 <a href="https://mcp-use.io/builder"><strong>Code Builder</strong></a></td>
+    <td>Explore MCP capabilities and generate starter code with the interactive <a href="https://mcp-use.io/builder">code builder</a>.</td>
+  </tr>
+  <tr>
+    <td>🔗 <a href="#http-connection-example"><strong>HTTP Support</strong></a></td>
+    <td>Direct connection to MCP servers running on specific HTTP ports</td>
+  </tr>
+  <tr>
+    <td>⚙️ <a href="#dynamic-server-selection-server-manager"><strong>Dynamic Server Selection</strong></a></td>
+    <td>Agents can dynamically choose the most appropriate MCP server for a given task from the available pool</td>
+  </tr>
+  <tr>
+    <td>🧩 <a href="#multi-server-support"><strong>Multi-Server Support</strong></a></td>
+    <td>Use multiple MCP servers simultaneously in a single agent</td>
+  </tr>
+  <tr>
+    <td>🛡️ <a href="#tool-access-control"><strong>Tool Restrictions</strong></a></td>
+    <td>Restrict potentially dangerous tools like file system or network access</td>
+  </tr>
+  <tr>
+    <td>🔧 <a href="#build-a-custom-agent"><strong>Custom Agents</strong></a></td>
+    <td>Build your own agents with any framework using the LangChain adapter or create new adapters</td>
+  </tr>
+  <tr>
+    <td>❓ <a href="https://mcp-use.io/what-should-we-build-next"><strong>What should we build next</strong></a></td>
+    <td>Let us know what you'd like us to build next</td>
+  </tr>
+</table>
 
 # Quick start
 
@@ -77,11 +111,8 @@ pip install langchain-openai
 
 # For Anthropic
 pip install langchain-anthropic
-
-# For other providers, check the [LangChain chat models documentation](https://python.langchain.com/docs/integrations/chat/)
 ```
-
-and add your API keys for the provider you want to use to your `.env` file.
+For other providers, check the [LangChain chat models documentation](https://python.langchain.com/docs/integrations/chat/) and add your API keys for the provider you want to use to your `.env` file.
 
 ```bash
 OPENAI_API_KEY=
@@ -520,6 +551,101 @@ if __name__ == "__main__":
     asyncio.run(main())
 ```
 
+# Sandboxed Execution
+
+MCP-Use supports running MCP servers in a sandboxed environment using E2B's cloud infrastructure. This allows you to run MCP servers without having to install dependencies locally, making it easier to use tools that might have complex setups or system requirements.
+
+## Installation
+
+To use sandboxed execution, you need to install the E2B dependency:
+
+```bash
+# Install mcp-use with E2B support
+pip install "mcp-use[e2b]"
+
+# Or install the dependency directly
+pip install e2b-code-interpreter
+```
+
+You'll also need an E2B API key. You can sign up at [e2b.dev](https://e2b.dev) to get your API key.
+
+## Configuration
+
+To enable sandboxed execution, use the `ClientOptions` parameter when creating your `MCPClient`:
+
+```python
+import asyncio
+import os
+from dotenv import load_dotenv
+from langchain_openai import ChatOpenAI
+from mcp_use import MCPAgent, MCPClient
+from mcp_use.types.sandbox import SandboxOptions
+from mcp_use.types.clientoptions import ClientOptions
+
+async def main():
+    # Load environment variables (needs E2B_API_KEY)
+    load_dotenv()
+
+    # Define MCP server configuration
+    server_config = {
+        "mcpServers": {
+            "everything": {
+                "command": "npx",
+                "args": ["-y", "@modelcontextprotocol/server-everything"],
+            }
+        }
+    }
+
+    # Define sandbox options
+    sandbox_options: SandboxOptions = {
+        "api_key": os.getenv("E2B_API_KEY"),  # API key can also be provided directly
+        "sandbox_template_id": "base",  # Use base template
+    }
+
+    # Create client options for sandboxed mode
+    client_options: ClientOptions = {
+        "is_sandboxed": True,
+        "sandbox_options": sandbox_options
+    }
+
+    # Create client with sandboxed mode enabled
+    client = MCPClient(
+        config=server_config,
+        options=client_options
+    )
+
+    # Create agent with the sandboxed client
+    llm = ChatOpenAI(model="gpt-4o")
+    agent = MCPAgent(llm=llm, client=client)
+
+    # Run your agent
+    result = await agent.run("Use the command line tools to help me add 1+1")
+    print(result)
+
+    # Clean up
+    await client.close_all_sessions()
+
+if __name__ == "__main__":
+    asyncio.run(main())
+```
+
+## Sandbox Options
+
+The `SandboxOptions` type provides configuration for the sandbox environment:
+
+| Option                 | Description                                                                              | Default               |
+| ---------------------- | ---------------------------------------------------------------------------------------- | --------------------- |
+| `api_key`              | E2B API key. Required - can be provided directly or via E2B_API_KEY environment variable | None                  |
+| `sandbox_template_id`  | Template ID for the sandbox environment                                                  | "base"                |
+| `supergateway_command` | Command to run supergateway                                                              | "npx -y supergateway" |
+
+## Benefits of Sandboxed Execution
+
+- **No local dependencies**: Run MCP servers without installing dependencies locally
+- **Isolation**: Execute code in a secure, isolated environment
+- **Consistent environment**: Ensure consistent behavior across different systems
+- **Resource efficiency**: Offload resource-intensive tasks to cloud infrastructure
+
 # Build a Custom Agent:
 
 You can also build your own custom agent using the LangChain adapter:
@@ -611,7 +737,6 @@ agent = MCPAgent(
 
 This is useful when you only need to see the agent's steps and decision-making process without all the low-level debug information from other components.
 
-
 # Roadmap
 
 <ul>

{mcp_use-1.2.13 → mcp_use-1.3.0}/docs/api-reference/introduction.mdx

@@ -49,6 +49,60 @@ client = MCPClient.from_dict(config=config)
 | --------- | ---- | -------- | ---------------------------------------------- |
 | `config`  | dict | Yes      | Dictionary containing MCP server configuration |
 
+#### Sandboxed Execution
+
+Both `from_config_file` and `from_dict` methods support the `options` parameter for configuring client features, including sandboxed execution:
+
+```python
+from mcp_use import MCPClient
+from mcp_use.types.sandbox import SandboxOptions
+from mcp_use.types.clientoptions import ClientOptions
+
+# Define sandbox options
+sandbox_options: SandboxOptions = {
+    "api_key": "your_e2b_api_key",
+    "sandbox_template_id": "code-interpreter-v1"
+}
+
+# Create client options for sandboxed mode
+client_options: ClientOptions = {
+    "is_sandboxed": True,
+    "sandbox_options": sandbox_options
+}
+
+# Create client with sandboxed mode enabled
+client = MCPClient.from_config_file(
+    config_path="config.json",
+    options=client_options
+)
+```
+
+| Parameter | Type          | Required | Default | Description                                          |
+| --------- | ------------- | -------- | ------- | ---------------------------------------------------- |
+| `options` | ClientOptions | No       | {}      | Client configuration options, including sandbox mode |
+
+The `ClientOptions` type is a TypedDict with the following fields:
+
+| Field             | Type           | Required | Default | Description                                       |
+| ----------------- | -------------- | -------- | ------- | ------------------------------------------------- |
+| `is_sandboxed`    | bool           | No       | False   | Whether to run commands in a sandbox environment  |
+| `sandbox_options` | SandboxOptions | No       | None    | Configuration options for the sandbox environment |
+
+The `SandboxOptions` type supports the following options:
+
+| Option                 | Type | Required | Default               | Description                                                                              |
+| ---------------------- | ---- | -------- | --------------------- | ---------------------------------------------------------------------------------------- |
+| `api_key`              | str  | Yes      | None                  | E2B API key. Required - can be provided directly or via E2B_API_KEY environment variable |
+| `sandbox_template_id`  | str  | No       | "base"                | Template ID for the sandbox environment                                                  |
+| `supergateway_command` | str  | No       | "npx -y supergateway" | Command to run supergateway                                                              |
+
+**When to use sandboxed execution**:
+
+- When you want to run MCP servers without installing their dependencies locally
+- To ensure consistent execution environments across different systems
+- For improved security through isolation
+- To leverage cloud resources for resource-intensive MCP servers
+
 ### Core Methods
 
 #### create_session
@@ -133,22 +187,22 @@ agent = MCPAgent(
 )
 ```
 
-| Parameter                 | Type                | Required | Default | Description                                                  |
-| ------------------------- | ------------------- | -------- | ------- | ------------------------------------------------------------ |
-| `llm`                     | BaseLanguageModel   | Yes      | -       | Any LangChain-compatible language model                      |
-| `client`                  | MCPClient           | No       | None    | The MCPClient instance                                       |
-| `connectors`              | list[BaseConnector] | No       | None    | List of connectors if not using client                       |
-| `server_name`             | str                 | No       | None    | Name of the server to use                                    |
-| `max_steps`               | int                 | No       | 5       | Maximum number of steps the agent can take                   |
-| `auto_initialize`         | bool                | No       | False   | Whether to initialize automatically                          |
-| `memory_enabled`          | bool                | No       | True    | Whether to enable memory                                     |
-| `system_prompt`           | str                 | No       | None    | Custom system prompt                                         |
-| `system_prompt_template`  | str                 | No       | None    | Custom system prompt template                                |
-| `additional_instructions` | str                 | No       | None    | Additional instructions for the agent                        |
-| `session_options`         | dict                | No       | {}      | Additional options for session creation                      |
-| `output_parser`           | OutputParser        | No       | None    | Custom output parser for LLM responses                       |
+| Parameter                 | Type                | Required | Default | Description                                                                                                                                     |
+| ------------------------- | ------------------- | -------- | ------- | ----------------------------------------------------------------------------------------------------------------------------------------------- |
+| `llm`                     | BaseLanguageModel   | Yes      | -       | Any LangChain-compatible language model                                                                                                         |
+| `client`                  | MCPClient           | No       | None    | The MCPClient instance                                                                                                                          |
+| `connectors`              | list[BaseConnector] | No       | None    | List of connectors if not using client                                                                                                          |
+| `server_name`             | str                 | No       | None    | Name of the server to use                                                                                                                       |
+| `max_steps`               | int                 | No       | 5       | Maximum number of steps the agent can take                                                                                                      |
+| `auto_initialize`         | bool                | No       | False   | Whether to initialize automatically                                                                                                             |
+| `memory_enabled`          | bool                | No       | True    | Whether to enable memory                                                                                                                        |
+| `system_prompt`           | str                 | No       | None    | Custom system prompt                                                                                                                            |
+| `system_prompt_template`  | str                 | No       | None    | Custom system prompt template                                                                                                                   |
+| `additional_instructions` | str                 | No       | None    | Additional instructions for the agent                                                                                                           |
+| `session_options`         | dict                | No       | {}      | Additional options for session creation                                                                                                         |
+| `output_parser`           | OutputParser        | No       | None    | Custom output parser for LLM responses                                                                                                          |
 | `use_server_manager`      | bool                | No       | False   | If `True`, enables automatic selection of the appropriate server based on the chosen tool when multiple servers are configured via `MCPClient`. |
-| `disallowed_tools`        | list[str]           | No       | None    | List of tool names that should not be available to the agent |
+| `disallowed_tools`        | list[str]           | No       | None    | List of tool names that should not be available to the agent                                                                                    |
 
 **When to use different parameters**:
 
@@ -459,3 +513,91 @@ This approach is useful when:
 - You want to limit the agent's capabilities for specific tasks
 - You're concerned about security implications of certain tools
 - You want to focus the agent on specific functionality
+
+### Sandboxed Execution with Multiple Servers
+
+Configure and use multiple sandboxed MCP servers:
+
+```python
+import os
+from dotenv import load_dotenv
+from mcp_use import MCPClient, MCPAgent
+from mcp_use.types.sandbox import SandboxOptions
+from langchain_anthropic import ChatAnthropic
+
+# Load environment variables
+load_dotenv()
+
+# Define sandbox options
+sandbox_options: SandboxOptions = {
+    "api_key": os.getenv("E2B_API_KEY"),
+    "sandbox_template_id": "code-interpreter-v1"
+}
+
+# Create client with multiple sandboxed servers
+client = MCPClient.from_dict(
+    {
+        "mcpServers": {
+            "browser": {
+                "command": "npx",
+                "args": ["@playwright/mcp@latest"]
+            },
+            "command": {
+                "command": "npx",
+                "args": ["-y", "@modelcontextprotocol/server-everything"]
+            }
+        }
+    },
+    is_sandboxed=True,
+    sandbox_options=sandbox_options
+)
+
+# Create agent with server manager enabled
+agent = MCPAgent(
+    llm=ChatAnthropic(model="claude-3-5-sonnet"),
+    client=client,
+    use_server_manager=True  # Automatically selects the appropriate server
+)
+
+# Run a task that will use tools from both servers
+result = await agent.run(
+    "Search for information about Python and then use the command line to check the latest version"
+)
+```
+
+This approach is useful when:
+
+- You need to use multiple MCP servers but don't want to install their dependencies locally
+- You want to ensure consistent execution environments for all servers
+- You need to leverage cloud resources for resource-intensive MCP servers
+
+## Error Handling
+
+mcp_use provides several exception types to handle different error scenarios:
+
+| Exception                | Description                       | When It Occurs                      |
+| ------------------------ | --------------------------------- | ----------------------------------- |
+| `MCPConnectionError`     | Connection to MCP server failed   | Network issues, server not running  |
+| `MCPAuthenticationError` | Authentication with server failed | Invalid credentials or tokens       |
+| `MCPTimeoutError`        | Operation timed out               | Server takes too long to respond    |
+| `MCPServerError`         | Server returned an error          | Internal server error               |
+| `MCPClientError`         | Client-side error                 | Invalid configuration or parameters |
+| `MCPError`               | Generic MCP-related error         | Any other MCP-related issue         |
+
+**Handling Strategies**:
+
+```python
+from mcp_use.exceptions import MCPConnectionError, MCPTimeoutError
+
+try:
+    result = await agent.run("Find information")
+except MCPConnectionError:
+    # Handle connection issues
+    print("Failed to connect to the MCP server")
+except MCPTimeoutError:
+    # Handle timeout issues
+    print("Operation timed out")
+except Exception as e:
+    # Handle other exceptions
+    print(f"An error occurred: {e}")
+```