Install Charles MCP Server

Pick your client, copy the command, done.

claude mcp add charles-mcp -- uvx charles-mcp

Environment Variables

Set these before running Charles MCP Server.

VariableDescriptionRequired

CHARLES_USERCharles Web Interface usernameNo

CHARLES_PASSCharles Web Interface passwordNo

CHARLES_PROXY_HOSTCharles proxy host addressNo

CHARLES_PROXY_PORTCharles proxy portNo

CHARLES_MANAGE_LIFECYCLEWhether the MCP server manages Charles process lifecycleNo

Once configured, Charles MCP Server gives your AI agent access to:

get_trafficRetrieves captured network traffic from Charles Proxy for analysis

filterlimit

get_request_detailFetches detailed information for a specific network request

request_id

After setup, try these prompts with your AI agent:

→Analyze the latest network traffic and identify any failed API requests.

→Show me a summary of the recent traffic captured by Charles Proxy.

→Get the full details for the request with ID 123 to debug the payload.

→Filter the current session traffic to show only requests to the production API.

Prerequisites & system requirements

An MCP-compatible client (Claude Code, Cursor, Windsurf, Claude Desktop, or Codex)
Python 3.8+ with pip installed

Alternative installation methods

Add to claude_desktop_config.json

Save the docs, env vars, and workflow around Charles MCP Server in Conare so Claude Code, Codex, and Cursor remember it next time.