Command Line Options

fast-agent offers flexible command line options for both running agent applications and using built-in CLI utilities.

Agent Applications

When running a fast-agent application (typically uv run agent.py), you have access to the following command line options:

Option	Description	Example
--model MODEL	Override the default model for the agent	--model gpt-4o
--agent AGENT	Specify which agent to use (default: "default")	--agent researcher
-m, --message MESSAGE	Send a single message to the agent and exit	--message "Hello world"
-p, --prompt-file FILE	Load and apply a prompt file	--prompt-file conversation.txt
--quiet	Disable progress display, tool and message logging	--quiet
--version	Show version and exit	--version
--server	Run as an MCP server	--server
--transport {sse,stdio}	Transport protocol when running as server	--transport sse
--port PORT	Port for SSE server (default: 8000)	--port 8080
--host HOST	Host for SSE server (default: 0.0.0.0)	--host localhost

Examples

# Run interactively with specified model
uv run agent.py --model sonnet

# Run specific agent
uv run agent.py --agent researcher

# Run with specific agent and model
uv run agent.py --agent researcher --model gpt-4o

# Send a message to an agent and exit
uv run agent.py --agent summarizer --message "Summarize this document"

# Apply a prompt file
uv run agent.py --prompt-file my_conversation.txt

# Run as an SSE server on port 8080
uv run agent.py --server --transport sse --port 8080

# Run as a stdio server
uv run agent.py --server --transport stdio

# Get minimal output (for scripting)
uv run agent.py --quiet --message "Generate a report"

Programmatic Control of Command Line Parsing

When embedding FastAgent in other applications (like web frameworks or GUI applications), you can disable command line parsing by setting parse_cli_args=False in the constructor:

# Create FastAgent without parsing command line arguments
fast = FastAgent("Embedded Agent", parse_cli_args=False)

This is particularly useful when: - Integrating with frameworks like FastAPI/Uvicorn that have their own argument parsing - Building GUI applications where command line arguments aren't relevant - Creating applications with custom argument parsing requirements

fast-agent go Command

The fast-agent go command lets you run an interactive agent directly without creating a Python file:

Option	Description	Example
--name NAME	Name for the agent (default: "FastAgent CLI")	--name "Research Assistant"
-i, --instruction INSTRUCTION	Instruction for the agent	--instruction "You are a coding assistant"
-c, --config-path PATH	Path to config file	--config-path ./config/fastagent.config.yaml
--servers SERVERS	Comma-separated list of server names	--servers fetch,filesystem
--model MODEL	Override the default model	--model sonnet
--quiet	Disable progress display and logging	--quiet

Examples

# Basic usage
fast-agent go

# Specify a model and instructions
fast-agent go --model haiku --instruction "You are a helpful AI assistant"

# Use specific MCP servers
fast-agent go --servers fetch,filesystem

# Specify a custom configuration file
fast-agent go --config-path ./my-config.yaml

# Create a named agent with specific servers and model
fast-agent go --name "Code Helper" --servers filesystem --model gpt-4o --instruction "You are a coding assistant specializing in Python"

fast-agent check Command

Use fast-agent check to diagnose your configuration:

# Show configuration summary
fast-agent check

# Display configuration file
fast-agent check show

# Display secrets file
fast-agent check show --secrets

fast-agent setup Command

Create a new agent project with configuration files:

# Set up in current directory
fast-agent setup

# Set up in a specific directory
fast-agent setup --config-dir ./my-agent

# Force overwrite existing files
fast-agent setup --force

fast-agent quickstart Command

Create example applications to get started quickly:

# Show available examples
fast-agent quickstart

# Create workflow examples
fast-agent quickstart workflow .

# Create researcher example
fast-agent quickstart researcher .

# Create data analysis example
fast-agent quickstart data-analysis .

# Create state transfer example
fast-agent quickstart state-transfer .