Skip to main content

​
Release Notes - mcp-use Python 1.4.1

​
Major Features

​
Code Mode Integration

A powerful new feature that enables direct code execution through MCP servers. This allows LLMs to write and execute code in isolated environments with full tool support. Key Capabilities:
  • Execute Python code in isolated environments
  • Full support for MCP tools within code execution
  • Enhanced security with sandboxed code execution
  • Seamless integration with existing MCPClient workflows
New Components:
  • CodeExecutor: Handles code execution with security and isolation
  • CodeModeConnector: New connector type for code mode operations
  • Integrated prompts for code mode interactions
Usage Example: 
from mcp_use import MCPClient
from mcp_use.client.code_executor import CodeExecutor

client = MCPClient()
# Enable code mode for your workflow
See full documentation at docs/python/client/code-mode.mdx

​
Pretty Display for Agent Interactions

Rich, colorful terminal output for agent operations, making it easier to follow agent thinking and tool calls in real-time. Features:
  • Beautiful formatted output for agent steps
  • Color-coded tool calls and responses
  • Structured display of thinking process
  • Enhanced readability for debugging
New Module:
  • mcp_use.agents.display: Rich console display utilities
Usage: 
from mcp_use.agents import MCPAgent

agent = MCPAgent(
    model=model,
    server_manager=server_manager,
    pretty_print=True  # Enable pretty display
)

​
Enhancements

​
SSL Certificate Verification Control

Added flexible SSL certificate verification control across HTTP connectors. Changes:
  • New verify parameter in MCPClient
  • SSL verification control in HttpConnector
  • Custom HTTPX client factory support in SSE and Streamable HTTP managers
Usage: 
client = MCPClient(verify=False)  # Disable SSL verification if needed

​
LangChain Integration

Comprehensive integration with LangChain for advanced agent workflows. Additions:
  • Full LangChain adapter support
  • Example implementation: langchain_integration_example.py
  • Documentation at docs/python/integration/langchain.mdx

​
Enhanced Schema Handling

Improved JSON schema processing for better compatibility with complex data structures. Fixes:
  • List handling in fix_schema method
  • Better type inference for nested structures
  • Improved adapter compatibility

​
Bug Fixes

​
Recursion Limit Configuration

Fixed issues with recursion limit configuration in MCPAgent to prevent stack overflow errors in deeply nested tool calls. Changes:
  • Proper handling of custom recursion limits
  • Better default configurations
  • Override protection

​
Client Info Property

Fixed the client_info property in base connector to properly return client metadata. Impact:
  • Correct client identification in server connections
  • Improved debugging capabilities

​
Code Quality

​
Import Structure Cleanup

Removed relative imports across the codebase for better maintainability and clarity. Files Updated:
  • mcp_use/client/connectors/http.py
  • mcp_use/client/connectors/sandbox.py
  • mcp_use/errors/error_formatting.py

​
Example Updates

Updated all example scripts to use latest model versions:
  • Claude: claude-3-5-sonnet-20240620 → claude-sonnet-4-5
  • OpenAI: gpt-4o → gpt-5

​
Dependencies

​
New Dependencies

  • rich>=14.2.0 - For pretty terminal display

​
Updated Requirements

  • langchain>=1.0.0 - Full LangChain 1.0 support
  • langchain-core>=1.0.0

​
Documentation

​
New Documentation

  • Code mode integration guide with examples
  • LangChain integration documentation
  • API reference updates for all new features

​
Updated Documentation

  • Client configuration with SSL verification options
  • Agent configuration with pretty print options
  • Connector documentation with new code mode connector

​
Breaking Changes

None in this release. All changes are backward compatible.

​
Migration Guide

​
Enabling Code Mode

To use the new code mode feature: 
from mcp_use import MCPClient
from mcp_use.client.connectors.code_mode import CodeModeConnector

client = MCPClient()
# Configure your code mode connector

​
Enabling Pretty Display

To enable pretty display in your agents: 
agent = MCPAgent(
    model=model,
    server_manager=server_manager,
    pretty_print=True
)

​
Disabling SSL Verification

If you need to disable SSL verification (e.g., for local development): 
client = MCPClient(verify=False)

​
Contributors

Thank you to all contributors who made this release possible:
  • Pietro Zullo (@pietrozullo)
  • Vincenzo Reina (@renvins)
  • Enrico Toniato (@tonxxd)

​
Statistics

  • Total Commits: 10 (Python-specific)
  • Files Changed: 50+
  • Lines Added: 2,500+
  • New Examples: 2
  • New Documentation Pages: 3
For full changelog and commit history, see the commit log.