Integrations

Drop-in templates for CI/CD and agent runtimes.

Platform Compatibility

Claude Desktop
Works

Claude Desktop

Point mcpServers to mcptrust proxy
Node.js
Works

Node MCP servers

Any local stdio-based server
Python
Works

Python MCP servers

Any Python MCP server
Docker
Works

Docker

mcptrust run supports docker run
LangChain
Works

Python Adapters

pip install langchain-mcptrust
AG
Works

Autogen Adapter

pip install autogen-mcptrust
CrewAI
Works

CrewAI Adapter

pip install crewai-mcptrust
Works

Generic Stdio

Any stdio-based MCP client/server
Anthropic

Claude Desktop Configuration

Configure Claude Desktop to use MCPTrust as a secure proxy.

{ "mcpServers": { "filesystem": { "command": "mcptrust", "args": [ "proxy", "--lock", "/absolute/path/to/mcp-lock.json", "--", "npx", "-y", "@modelcontextprotocol/server-filesystem", "/tmp" ] } } }
Python

Python Agent (LangChain)

Wrap your server execution in Python agents.

from mcptrust_core import MCPTrust from langchain_mcptrust import TrustedMCPServer # Initialize trust engine mcp = MCPTrust() # Create a trusted server wrapper server = TrustedMCPServer( mcp=mcp, server_command="python mcp_server.py", preset="baseline", ) # Verify before use server.check() # Use in your agent tools = server.get_tools()
Docker

Docker Integration

Lock and run Docker-based MCP servers.

# Lock a Docker-based server mcptrust lock -- docker run -i mcp-server-image # Run with enforcement (Verification + Execution) mcptrust run --lock mcp-lock.json # Proxy mode (Advanced) mcptrust proxy --lock mcp-lock.json -- docker run -i mcp-server-image

GitHub Actions (Ed25519)

Add this workflow to .github/workflows/mcptrust.yml to block pull requests that drift from the lockfile.

name: MCPTrust Verification on: [pull_request] jobs: verify: runs-on: ubuntu-latest steps: - uses: actions/checkout@v4 - uses: mcptrust/mcptrust/.github/actions/mcptrust@v0.1.1 with: mode: check lockfile: mcp-lock.json fail_on: critical server_command: 'npx -y @modelcontextprotocol/server-filesystem /tmp' policy: baseline verify_provenance: true report: pr_comment