CLI Client

# Natural Language Agent Chat Client Configuration Example
#
# Configuration Priority (highest to lowest):
#   1. Command line flags
#   2. Environment variables
#   3. Configuration file values (this file)
#   4. Hard-coded defaults
#
# Copy this file to .pgedge-pg-mcp-cli.yaml or ~/.pgedge-pg-mcp-cli.yaml
# and customize it for your environment.

# ============================================================================
# MCP SERVER CONNECTION CONFIGURATION
# ============================================================================
mcp:
    # Connection mode: "stdio" or "http"
    # stdio: Spawns MCP server as subprocess (local only)
    # http: Connects to MCP server via HTTP/HTTPS (can be remote)
    # Default: stdio
    # Environment variable: PGEDGE_MCP_MODE
    # Command line flag: -mcp-mode
    mode: stdio

    # -------------------------
    # Stdio Mode Configuration
    # -------------------------
    # Path to MCP server binary (for stdio mode)
    # Default: ./bin/pgedge-postgres-mcp
    # Environment variable: PGEDGE_MCP_SERVER_PATH
    # Command line flag: -mcp-server-path
    server_path: ./bin/pgedge-postgres-mcp

    # Path to MCP server config file (for stdio mode)
    # If specified, the --config flag will be passed to the server binary
    # If not specified, the server will use its default config file lookup
    # Default: (none - server uses default config lookup)
    # Environment variable: PGEDGE_MCP_SERVER_CONFIG_PATH
    server_config_path: ./bin/pgedge-postgres-mcp-stdio.yaml

    # -------------------------
    # HTTP Mode Configuration
    # -------------------------
    # MCP server URL (for HTTP mode)
    # Should include protocol and optionally path: http://host:port or https://host:port
    # The /mcp/v1 path will be appended automatically if not present
    # Default: http://localhost:8080
    # Environment variable: PGEDGE_MCP_URL
    # Command line flag: -mcp-url
    # url: http://localhost:8080

    # Authentication mode for HTTP mode
    # Options:
    #   none:  No authentication (for servers with auth disabled)
    #   token: API token authentication (default)
    #   user:  Username/password authentication
    # Default: user (will prompt for credentials)
    # Environment variable: PGEDGE_MCP_AUTH_MODE
    # Command line flag: -mcp-auth-mode
    # auth_mode: user

    # Authentication token (for token auth mode)
    # Token priority: PGEDGE_MCP_TOKEN env var >
    #                 ~/.pgedge-pg-mcp-cli-token file >
    #                 this config value >
    #                 prompt at startup
    # Environment variable: PGEDGE_MCP_TOKEN
    # Token file: ~/.pgedge-pg-mcp-cli-token
    # Command line flag: -mcp-token
    # token: your-token-here

    # Username for HTTP mode (for user auth mode)
    # Environment variable: PGEDGE_MCP_USERNAME
    # Command line flag: -mcp-username
    # username: your-username

    # Password for HTTP mode (for user auth mode)
    # Environment variable: PGEDGE_MCP_PASSWORD
    # Command line flag: -mcp-password
    # password: your-password

    # Use TLS/HTTPS for HTTP mode
    # Default: false
    # Command line flag: (inferred from URL protocol)
    # tls: false

# ============================================================================
# LLM PROVIDER CONFIGURATION
# ============================================================================
llm:
    # Provider: "anthropic", "openai", or "ollama"
    # anthropic: Uses Anthropic's Claude API (requires API key)
    # openai: Uses OpenAI's GPT API (requires API key)
    # ollama: Uses locally running Ollama server (no API key needed)
    # Default: anthropic
    # Environment variable: PGEDGE_LLM_PROVIDER
    # Command line flag: -llm-provider
    provider: anthropic

    # Model to use
    # Anthropic models: claude-sonnet-4-20250514, claude-opus-4-20250514, etc.
    # OpenAI models: gpt-5-main, gpt-5-thinking, gpt-4o, gpt-4-turbo, gpt-3.5-turbo, etc.
    # Ollama models: llama3, llama3.1, mistral, gpt-oss:20b, etc.
    # Default: claude-sonnet-4-20250514 (anthropic), gpt-5-main (openai), or llama3 (ollama)
    # Environment variable: PGEDGE_LLM_MODEL
    # Command line flag: -llm-model
    model: claude-sonnet-4-20250514

    # -------------------------
    # Anthropic Configuration
    # -------------------------
    # API key for Anthropic
    # Get your API key from: https://console.anthropic.com/
    #
    # Priority (highest to lowest):
    # 1. Environment variable: PGEDGE_ANTHROPIC_API_KEY or ANTHROPIC_API_KEY
    # 2. API key file: anthropic_api_key_file
    # 3. Direct config value: anthropic_api_key (not recommended)
    #
    # Command line flag: -anthropic-api-key
    #
    # Option 1: Environment variable (recommended for development)
    # export PGEDGE_ANTHROPIC_API_KEY="sk-ant-your-key-here"
    #
    # Option 2: API key file (recommended for production)
    anthropic_api_key_file: ~/.anthropic-api-key
    #
    # Option 3: Direct value (not recommended - use env var or file)
    # anthropic_api_key: your-anthropic-api-key-here

    # -------------------------
    # OpenAI Configuration
    # -------------------------
    # API key for OpenAI
    # Get your API key from: https://platform.openai.com/
    #
    # Priority (highest to lowest):
    # 1. Environment variable: PGEDGE_OPENAI_API_KEY or OPENAI_API_KEY
    # 2. API key file: openai_api_key_file
    # 3. Direct config value: openai_api_key (not recommended)
    #
    # Command line flag: -openai-api-key
    #
    # Option 1: Environment variable (recommended for development)
    # export PGEDGE_OPENAI_API_KEY="sk-proj-your-key-here"
    #
    # Option 2: API key file (recommended for production)
    openai_api_key_file: ~/.openai-api-key
    #
    # Option 3: Direct value (not recommended - use env var or file)
    # openai_api_key: your-openai-api-key-here

    # Maximum tokens for LLM response
    # For GPT-5 and o-series models, automatically uses max_completion_tokens
    # For older models, uses max_tokens
    # Default: 4096
    # Command line flag: (not available)
    max_tokens: 4096

    # Temperature for sampling (0.0-1.0)
    # Lower = more focused/deterministic, Higher = more creative/random
    # Note: GPT-5 and o-series models only support default temperature (1)
    # Default: 0.7
    # Command line flag: (not available)
    temperature: 0.7

    # -------------------------
    # Ollama Configuration
    # -------------------------
    # Ollama server URL
    # Default: http://localhost:11434
    # Environment variable: PGEDGE_OLLAMA_URL
    # Command line flag: -ollama-url
    ollama_url: http://localhost:11434

# ============================================================================
# USER INTERFACE CONFIGURATION
# ============================================================================
ui:
    # Disable colored output
    # Useful for environments that don't support ANSI color codes
    # Can be toggled at runtime with /set color <on|off>
    # Note: NO_COLOR environment variable takes precedence if set
    # Default: false
    # Command line flag: -no-color
    no_color: false

    # Display status messages during tool execution
    # Shows messages like "→ Executing tool: query_database" during operations
    # Can be toggled at runtime with /set status-messages <on|off>
    # Default: true
    # Command line flag: (not available, use /set command at runtime)
    display_status_messages: true

    # Enable markdown rendering in assistant responses
    # Can be toggled at runtime with /set markdown <on|off>
    # Default: true
    # Command line flag: (not available, use /set command at runtime)
    render_markdown: true

Configuration Examples

Stdio Mode with Anthropic Claude

mcp:
    mode: stdio
    server_path: ./bin/pgedge-postgres-mcp
    server_config_path: ./bin/pgedge-postgres-mcp-stdio.yaml

llm:
    provider: anthropic
    model: claude-sonnet-4-20250514
    # Set PGEDGE_ANTHROPIC_API_KEY environment variable

Then run:

export PGEDGE_ANTHROPIC_API_KEY="your-key-here"
./bin/pgedge-nla-cli

Stdio Mode with OpenAI

mcp:
    mode: stdio
    server_path: ./bin/pgedge-postgres-mcp
    server_config_path: ./bin/pgedge-postgres-mcp-stdio.yaml

llm:
    provider: openai
    model: gpt-5-main
    # Set PGEDGE_OPENAI_API_KEY environment variable

Then run:

export PGEDGE_OPENAI_API_KEY="your-key-here"
./bin/pgedge-nla-cli

HTTP Mode with Token Authentication

mcp:
    mode: http
    url: https://mcp.example.com:8080
    auth_mode: token
    tls: true

llm:
    provider: anthropic
    model: claude-sonnet-4-20250514

Then run:

export PGEDGE_ANTHROPIC_API_KEY="your-key-here"
export PGEDGE_MCP_TOKEN="your-mcp-token"
./bin/pgedge-nla-cli

HTTP Mode without Authentication

For servers with authentication disabled:

mcp:
    mode: http
    url: http://localhost:8080
    auth_mode: none

llm:
    provider: anthropic
    model: claude-sonnet-4-20250514

Then run:

export PGEDGE_ANTHROPIC_API_KEY="your-key-here"
./bin/pgedge-nla-cli

Local Setup with Ollama

mcp:
    mode: stdio
    server_path: ./bin/pgedge-postgres-mcp
    server_config_path: ./bin/pgedge-postgres-mcp-stdio.yaml

llm:
    provider: ollama
    model: llama3
    ollama_url: http://localhost:11434

Then run:

# Make sure Ollama is running with the model pulled
ollama pull llama3
./bin/pgedge-nla-cli

Remote HTTP Server with Ollama

mcp:
    mode: http
    url: http://mcp-server.internal:8080

llm:
    provider: ollama
    model: gpt-oss:20b
    ollama_url: http://localhost:11434

Environment Variables

The chat client supports the following environment variables (in order of precedence):

MCP Connection

PGEDGE_MCP_MODE: Connection mode (stdio or http)
PGEDGE_MCP_SERVER_PATH: Path to MCP server binary (stdio mode)
PGEDGE_MCP_SERVER_CONFIG_PATH: Path to MCP server config file (stdio mode)
PGEDGE_MCP_URL: MCP server URL (http mode)
PGEDGE_MCP_AUTH_MODE: Authentication mode (none, token, or user)
PGEDGE_MCP_TOKEN: Authentication token (for token auth mode)
PGEDGE_MCP_USERNAME: Username (for user auth mode)
PGEDGE_MCP_PASSWORD: Password (for user auth mode)

LLM Configuration

PGEDGE_LLM_PROVIDER: LLM provider (anthropic, openai, or ollama)
PGEDGE_LLM_MODEL: Model to use
PGEDGE_ANTHROPIC_API_KEY: Anthropic API key
PGEDGE_OPENAI_API_KEY: OpenAI API key
PGEDGE_OLLAMA_URL: Ollama server URL

Command Line Flags

All configuration options can be overridden with command line flags:

./bin/pgedge-nla-cli \
    -config /path/to/config.yaml \
    -mcp-mode http \
    -mcp-url https://mcp.example.com:8080 \
    -mcp-auth-mode token \
    -mcp-token your-mcp-token \
    -llm-provider anthropic \
    -llm-model claude-opus-4-20250514 \
    -anthropic-api-key your-anthropic-key \
    -no-color

Available MCP authentication flags:

-mcp-auth-mode: Authentication mode (none, token, or user)
-mcp-token: Authentication token (for token mode)
-mcp-username: Username (for user mode)
-mcp-password: Password (for user mode)

Run ./bin/pgedge-nla-cli --help to see all available flags.

Token File Location

For HTTP mode authentication, the token can be stored in:

~/.pgedge-pg-mcp-cli-token

This file should contain only the token (no newlines or extra whitespace).

CLI Client

Configuration Examples

Stdio Mode with Anthropic Claude

Stdio Mode with OpenAI

HTTP Mode with Token Authentication

HTTP Mode without Authentication

Local Setup with Ollama

Remote HTTP Server with Ollama

Environment Variables

MCP Connection

LLM Configuration

Command Line Flags

Token File Location

See Also