HTTP MCP Transport for Nchan - Python SDK

This is an HTTP-based MCP (Machine Conversation Protocol) transport library designed for integration with Nchan.

Installation

pip install httmcp

Usage

from httmcp import HTTMCP

# Create MCP server
mcp_server = HTTMCP(
    name="my-mcp",
    instructions="This is an MCP server",
    publish_server="http://localhost:8080"
)

# Add MCP server to FastAPI application
app = FastAPI()
app.include_router(mcp_server.router)

OpenAPI Support

HTTMCP also supports creating MCP servers from OpenAPI specifications:

from httmcp import OpenAPIMCP

# Create MCP server from OpenAPI specification
mcp_server = await OpenAPIMCP.from_openapi(
    definition="openapi.json",
    name="my-openapi-mcp",
    publish_server="http://localhost:8080"
)

# Add MCP server to FastAPI application
app = FastAPI()
app.include_router(mcp_server.router)

One-Click GPTs Actions to MCP Deployment

HTTMCP provides a powerful CLI for instant deployment of GPTs Actions to MCP servers:

# Installation
pip install httmcp[cli]

# One-click deployment from GPTs Actions OpenAPI spec
python -m httmcp -f gpt_actions_openapi.yml -p http://nchan:80

# Complete deployment with custom configuration
python -m httmcp \
  --openapi-file gpt_actions_openapi.yml \
  --name "my-gpts-service" \
  --publish-server http://nchan:80 \
  --host 0.0.0.0 \
  --port 8080

CLI options:

• -f, --openapi-file: OpenAPI specification file path or URL (required)
• -p, --publish-server: Nchan publish server URL for MCP transport (required)
• -n, --name: Custom name for your MCP server (default: derived from OpenAPI)
• -H, --host: Host address to bind the server (default: 0.0.0.0)
• -P, --port: Local port number for accessing your deployed MCP service (default: 8000)