This document describes the configuration format for the MCP Client CLI. The configuration file uses JSON format (with support for comments via the commentjson
library).
The configuration file can be placed in either:
~/.llm/config.json
(user's home directory)mcp-server-config.json
(in the current working directory)
{
"systemPrompt": "string",
"llm": {
"provider": "string",
"model": "string",
"api_key": "string",
"temperature": float,
"base_url": "string"
},
"mcpServers": {
"server_name": {
"command": "string",
"args": ["string"],
"env": {
"ENV_VAR_NAME": "value"
},
"enabled": boolean,
"exclude_tools": ["string"],
"requires_confirmation": ["string"]
}
}
}
Field | Type | Required | Description |
---|---|---|---|
systemPrompt |
string | Yes | System prompt for the LLM |
llm |
object | No | LLM configuration |
mcpServers |
object | Yes | Dictionary of MCP server configurations |
Field | Type | Required | Default | Description |
---|---|---|---|---|
provider |
string | No | "openai" |
LLM provider |
model |
string | No | "gpt-4o" |
LLM model name |
api_key |
string | No | Environment vars | API key for the LLM service |
temperature |
float | No | 0 |
Temperature for LLM responses |
base_url |
string | No | null |
Custom API endpoint URL |
Notes:
- The
api_key
can be omitted if it's set via environment variablesLLM_API_KEY
orOPENAI_API_KEY
Field | Type | Required | Default | Description |
---|---|---|---|---|
command |
string | Yes | - | Command to run the server |
args |
array | No | [] |
Command-line arguments |
env |
object | No | {} |
Environment variables |
enabled |
boolean | No | true |
Whether the server is enabled |
exclude_tools |
array | No | [] |
Tool names to exclude |
requires_confirmation |
array | No | [] |
Tools requiring user confirmation |
{
"systemPrompt": "You are an AI assistant helping a software engineer...",
"llm": {
"provider": "openai",
"model": "gpt-4o-mini",
"api_key": "your-api-key-here",
"temperature": 0
},
"mcpServers": {
"fetch": {
"command": "uvx",
"args": ["mcp-server-fetch"]
},
"brave-search": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-brave-search"],
"env": {
"BRAVE_API_KEY": "your-brave-api-key-here"
}
},
"mcp-server-commands": {
"command": "npx",
"args": ["mcp-server-commands"],
"requires_confirmation": [
"run_command",
"run_script"
]
}
}
}
The configuration file supports comments with //
syntax:
{
"systemPrompt": "You are an AI assistant helping a software engineer...",
// Uncomment this section to use Anthropic Claude
// "llm": {
// "provider": "anthropic",
// "model": "claude-3-opus-20240229",
// "api_key": "your-anthropic-api-key"
// },
"llm": {
"provider": "openai",
"model": "gpt-4o",
"api_key": "your-openai-api-key"
}
}