The Agent Stack command-line interface (CLI) provides tools to manage agents, configure models, and control the platform directly from your terminal.
Basic Usage
agentstack [COMMAND] [OPTIONS] [ARGUMENTS]
To see all available commands:
Agent Commands
list
List all available agents:
Shows agent names, statuses, descriptions, locations, and any configuration issues.
run
Run an agent with specified input:
agentstack run <agent-name> [input]
Examples:
# Interactive mode
agentstack run chat
# Direct input
agentstack run chat "Hello!"
# Input from file
agentstack run chat < input.txt
# Output to file
agentstack run chat "Summarize this" > output.txt
# Append to file
agentstack run chat "More analysis" >> output.txt
# Pipe between agents
agentstack run agent1 "input" | agentstack run agent2
info
Get detailed information about a specific agent:
agentstack info <agent-name>
Shows description, input/output schemas, examples, and configuration options.
add
Add an agent to your environment:
agentstack add <location>
Location formats:
# From Docker image
agentstack add ghcr.io/i-am-bee/agentstack/agents/chat:latest
# From GitHub repository
agentstack add "https://github.com/user/repo"
# From GitHub with version tag
agentstack add "https://github.com/user/repo#v1.0.0"
# From GitHub with specific path
agentstack add "https://github.com/user/repo#:agents/my-agent"
# From local directory (builds Docker image)
agentstack add ./my-agent-directory
remove
Remove an agent from your environment:
agentstack remove <agent-name>
logs
View and stream logs for an agent:
agentstack logs <agent-name>
Displays the agent’s log output and continues streaming new logs as they are generated.
build
Build an agent from a local directory or GitHub repository:
agentstack build [context]
Arguments:
context: Path to build context (default: current directory)
Options:
--dockerfile <path>: Use custom dockerfile path
--tag <name>: Docker tag for the agent
--multi-platform: Build for multiple platforms
--push: Push the image to the target registry
--import/--no-import: Import the image into Agent Stack (default: true)
Examples:
# Build from current directory
agentstack build .
# Build with custom tag
agentstack build . --tag my-agent:latest
# Build without importing to platform
agentstack build . --no-import
# Build from GitHub URL
agentstack build "https://github.com/user/repo"
# Build with custom dockerfile
agentstack build . --dockerfile ./custom.Dockerfile
server-side-build
EXPERIMENTAL: Build agent from GitHub repository in the platform:
agentstack server-side-build <github-url>
This builds the agent on the server side rather than locally.
Model Commands
model setup
Interactive setup for LLM provider configuration:
Guides you through configuring your LLM provider with the required API keys and settings.
start
Starts the platform:
agentstack platform start
stop
Stops the platform:
Stops the platform but keeps all data intact.
delete
Deletes the platform and all data:
agentstack platform delete
This removes all agents, configurations, and data permanently.
MCP Commands
mcp
Manage MCP servers and toolkits:
agentstack mcp [subcommand]
Use agentstack mcp --help for available subcommands.
Server Commands
server
Manage Agent Stack servers and authentication:
login
Login to a server or switch between logged in servers:
logout
Logout from a server:
show
Show current server information:
list
List all configured servers:
Environment Commands
env
Manage environment variables:
agentstack env [subcommand]
Use agentstack env --help for available subcommands.
System Commands
Launch the web interface:
Opens Agent Stack’s graphical interface in your default browser.
version
Display version information:
install
Install Agent Stack CLI and platform:
upgrade
Upgrade Agent Stack CLI and platform to the latest version:
uninstall
Uninstall Agent Stack CLI and platform:
agentstack self uninstall
Removes the CLI, stops the platform, and deletes all data.
Global Options
Most commands support these options:
--help: Show help information for a command
-v, --verbose: Show verbose output
--debug: Enable debug output for troubleshooting (set via environment)