### Install MCP Executor with Smithery CLI

Source: https://context7.com/maxim-saplin/mcp_safe_local_python_executor/llms.txt

Use the Smithery CLI to automatically install and configure the MCP executor for Claude Desktop. This command simplifies the setup process.

```bash
# --- Smithery auto-install (installs and configures for Claude Desktop) ---
npx -y @smithery/cli install @maxim-saplin/mcp_safe_local_python_executor --client claude
```

--------------------------------

### Setup Development Environment with uv

Source: https://context7.com/maxim-saplin/mcp_safe_local_python_executor/llms.txt

Create a virtual environment using `uv` and install all project dependencies, including the development group. This is the initial step for local development and testing.

```bash
# Create virtualenv and install all dependencies (including dev group)
uv venv .venv
uv sync --group dev
```

--------------------------------

### Development Setup and Testing

Source: https://github.com/maxim-saplin/mcp_safe_local_python_executor/blob/main/README.md

Commands for setting up the development environment, installing dependencies, and running tests using `uv` and `pytest`.

```bash
uv venv .venv
uv sync --group dev
python -m pytest tests/
```

--------------------------------

### Build and Run MCP Executor with Docker

Source: https://context7.com/maxim-saplin/mcp_safe_local_python_executor/llms.txt

Build a Docker image for the safe-python-executor and run it. This method provides a pre-built environment with all dependencies installed, enabling offline container-based deployments.

```bash
# --- Docker (pre-built venv, no runtime downloads) ---
docker build -t safe-python-executor .
docker run --rm -i safe-python-executor
# Runs: uv run mcp_server.py  (CMD in Dockerfile)
```

--------------------------------

### Install Safe Local Python Executor via Smithery

Source: https://github.com/maxim-saplin/mcp_safe_local_python_executor/blob/main/README.md

Use the Smithery CLI to automatically install the Safe Local Python Executor for Claude Desktop.

```bash
npx -y @smithery/cli install @maxim-saplin/mcp_safe_local_python_executor --client claude
```

--------------------------------

### Configure MCP Server for Claude Desktop

Source: https://context7.com/maxim-saplin/mcp_safe_local_python_executor/llms.txt

Configure Claude Desktop to use the safe-local-python-executor by specifying the command and arguments in `claude_desktop_config.json`. Ensure the path to the repository is absolute.

```json
# --- Claude Desktop configuration (claude_desktop_config.json) ---
# macOS:  ~/Library/Application Support/Claude/claude_desktop_config.json
# Windows: %APPDATA%\Claude\claude_desktop_config.json
{
    "mcpServers": {
        "safe-local-python-executor": {
            "command": "uv",
            "args": [
                "--directory",
                "/absolute/path/to/mcp_safe_local_python_executor/",
                "run",
                "mcp_server.py"
            ]
        }
    }
}
```

--------------------------------

### Launch MCP Server with Stdio Transport

Source: https://context7.com/maxim-saplin/mcp_safe_local_python_executor/llms.txt

Manually launch the MCP server using `uv` from the repository root. This is the standard transport for MCP tool servers integrated with desktop AI clients.

```bash
# --- Manual launch (from repo root) ---
uv run mcp_server.py
# 2025-01-01 00:00:00 - root - INFO - Starting MCP server for Python executor
```

--------------------------------

### Configure Claude Desktop MCP Servers

Source: https://github.com/maxim-saplin/mcp_safe_local_python_executor/blob/main/README.md

Add the Safe Local Python Executor configuration to your Claude Desktop settings to enable it as an MCP server.

```json
{
    "mcpServers": {
        "safe-local-python-executor": {
            "command": "uv",
            "args": [
                "--directory", 
                "/path/to/mcp_local_python_executor/",
                "run",
                "mcp_server.py"
            ]
        }
    }
}
```

--------------------------------

### Run Full Test Suite with pytest

Source: https://context7.com/maxim-saplin/mcp_safe_local_python_executor/llms.txt

Execute the entire test suite using `pytest` from the command line. This command runs tests in both `test_local_python_executor.py` and `test_mcp_server.py`.

```bash
# Run the full test suite
python -m pytest tests/ -v
```

--------------------------------

### run_python - Execute sandboxed Python code via MCP

Source: https://context7.com/maxim-saplin/mcp_safe_local_python_executor/llms.txt

The primary MCP tool exposed by the server. It accepts a string of valid Python 3 code, runs it inside the `LocalPythonExecutor` sandbox, and returns a dictionary containing `result` and `logs`. The sandbox enforces a fixed import allowlist and blocks all file-system, network, and subprocess operations.

```APIDOC
## run_python

### Description
Executes sandboxed Python code and returns the result and any print output.

### Method
MCP Tool Call

### Parameters
#### Request Body
- **code** (string) - Required - The Python code to execute.

### Response
#### Success Response (200)
- **result** (any) - The computed value of the last expression or the `result` variable.
- **logs** (string) - All captured `print` output.

### Request Example
```python
import asyncio
import sys, os
sys.path.insert(0, os.path.abspath("..\'))  # point to repo root
import mcp_server

response = asyncio.run(mcp_server.run_python("result = 2 + 2"))
print(response)
```

### Response Example
```json
{
  "result": 4,
  "logs": ""
}
```

### Error Handling
- **InterpreterError**: Raised if the code attempts to perform disallowed operations (e.g., unauthorized imports, file I/O, network access, subprocesses).

### Request Example (Blocked Import)
```python
import pytest
import asyncio
import sys, os
sys.path.insert(0, os.path.abspath("..\'))  # point to repo root

with pytest.raises(Exception) as exc_info:
    asyncio.run(mcp_server.run_python("import os\nresult = os.listdir('.')"))
print(exc_info.value)
```

### Response Example (Blocked Import)
```
InterpreterError: Import of os is not allowed. ...
```
```

--------------------------------

### Run Specific Test File with pytest

Source: https://context7.com/maxim-saplin/mcp_safe_local_python_executor/llms.txt

Execute tests from a specific file, `test_mcp_server.py`, using `pytest`. This is useful for focusing on a particular set of tests.

```bash
# Run a specific test file
python -m pytest tests/test_mcp_server.py -v
```

--------------------------------

### Execute Python Code via MCP Tool

Source: https://context7.com/maxim-saplin/mcp_safe_local_python_executor/llms.txt

Use the `run_python` MCP tool to execute Python code in a sandboxed environment. It returns the computed result and any print output. Ensure code produces a `result` variable or ends with an expression for a meaningful return value.

```python
# Called internally by the MCP framework; direct async usage from tests:
import asyncio
import sys, os
sys.path.insert(0, os.path.abspath(".."))  # point to repo root
import mcp_server

# --- Basic arithmetic ---
response = asyncio.run(mcp_server.run_python("result = 2 + 2"))
print(response)
# {"result": 4, "logs": ""}
```

```python
# --- Using an allowed stdlib module ---
response = asyncio.run(mcp_server.run_python(
    "import math\nresult = math.sqrt(16)"
))
print(response)
# {"result": 4.0, "logs": ""}
```

```python
# --- Capturing print output ---
response = asyncio.run(mcp_server.run_python(
    "print('hello')\nresult = 42"
))
print(response)
# {"result": 42, "logs": "hello\n"}
```

```python
# --- Complex computation: primes up to 100 ---
code = """
def is_prime(n):
    if n < 2:
        return False
    for i in range(2, int(n**0.5) + 1):
        if n % i == 0:
            return False
    return True

primes = [num for num in range(2, 101) if is_prime(num)]
result = primes
"""
response = asyncio.run(mcp_server.run_python(code))
print(response["result"])
# [2, 3, 5, 7, 11, 13, 17, 19, 23, 29, 31, 37, 41,
#  43, 47, 53, 59, 61, 67, 71, 73, 79, 83, 89, 97]
```

```python
# --- Blocked import raises an exception ---
import pytest
with pytest.raises(Exception) as exc_info:
    asyncio.run(mcp_server.run_python("import os\nresult = os.listdir('.')"))
print(exc_info.value)
# InterpreterError: Import of os is not allowed. ...
```

--------------------------------

### Low-level Sandboxed Python Interpreter

Source: https://context7.com/maxim-saplin/mcp_safe_local_python_executor/llms.txt

Utilize `LocalPythonExecutor` for direct sandboxed Python execution. It returns the result, logs, and a third unused value. The sandbox restricts imports, file I/O, and network access.

```python
from smolagents.local_python_executor import LocalPythonExecutor

# Instantiate with no extra imports
executor = LocalPythonExecutor(additional_authorized_imports=[])
executor.send_tools({})  # no external tools

# Simple expression — last expression value is the result
result, logs, _ = executor("2 + 2")
assert result == 4
assert logs == ""
```

```python
# Variable assignment — explicit `result` variable
result, logs, _ = executor("""
x = 10
y = 20
result = x + y
""")
assert result == 30
```

```python
# List iteration
result, logs, _ = executor("""
numbers = [1, 2, 3, 4, 5]
total = 0
for num in numbers:
    total += num
total
""")
assert result == 15
```

```python
# Blocked import — raises InterpreterError
import pytest
with pytest.raises(Exception) as exc_info:
    executor("import os\nos.listdir('.')")
assert "Import of os is not allowed" in str(exc_info.value)
```

--------------------------------

### LocalPythonExecutor - Low-level sandboxed interpreter

Source: https://context7.com/maxim-saplin/mcp_safe_local_python_executor/llms.txt

The underlying execution engine from `smolagents`. It is callable and executes Python code within a restricted sandbox, blocking unauthorized imports and file-system operations.

```APIDOC
## LocalPythonExecutor

### Description
Provides a low-level interface to the sandboxed Python interpreter.

### Method
Instantiate and call the executor object.

### Parameters
#### Initialization
- **additional_authorized_imports** (list of strings) - Optional - A list of additional modules allowed to be imported. Defaults to an empty list, using only the built-in allowlist.

#### Execution
- **code** (string) - Required - The Python code to execute.

### Request Example (Instantiation)
```python
from smolagents.local_python_executor import LocalPythonExecutor

executor = LocalPythonExecutor(additional_authorized_imports=[])
executor.send_tools({})
```

### Response
#### Success Response (Execution)
- **result** (any) - The computed value of the last expression or the `result` variable.
- **logs** (string) - All captured `print` output.
- **_** (any) - An unused third return value.

### Request Example (Simple Arithmetic)
```python
result, logs, _ = executor("2 + 2")
assert result == 4
assert logs == ""
```

### Response Example (Simple Arithmetic)
```json
{
  "result": 4,
  "logs": ""
}
```

### Request Example (Variable Assignment)
```python
result, logs, _ = executor("\nx = 10\ny = 20\nresult = x + y\n")
assert result == 30
```

### Response Example (Variable Assignment)
```json
{
  "result": 30,
  "logs": ""
}
```

### Error Handling
- **InterpreterError**: Raised if the code attempts to perform disallowed operations (e.g., unauthorized imports, file I/O, network access, subprocesses).

### Request Example (Blocked Import)
```python
import pytest

with pytest.raises(Exception) as exc_info:
    executor("import os\nos.listdir('.')")
assert "Import of os is not allowed" in str(exc_info.value)
```
```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.