Struct Launcher 

pub struct Launcher { /* private fields */ }

Available on crate feature cli only.

CLI launcher for running agents.

Quick way to run agents in console or server mode:

• Launcher - Main entry point for CLI apps

Available with feature: cli Launcher for running ADK agents with CLI support.

Provides console and web server modes out of the box.

Console mode

Uses a rustyline REPL with history, Ctrl+C handling, and streaming output that renders <think> blocks, tool calls, and inline data.

Serve mode

Starts an Axum HTTP server with the ADK web UI.

Note on with_streaming_mode

with_streaming_mode currently only affects console mode. ServerConfig does not accept a RunConfig, so the setting is not forwarded to the server. This will be addressed when ServerConfig gains that field.

Implementations

impl Launcher

pub fn new(agent: Arc<dyn Agent>) -> Launcher

Create a new launcher with the given agent.

pub fn app_name(self, name: impl Into<String>) -> Launcher

Set a custom application name (defaults to agent name).

pub fn with_artifact_service( self, service: Arc<dyn ArtifactService>, ) -> Launcher

Set a custom artifact service.

pub fn with_session_service(self, service: Arc<dyn SessionService>) -> Launcher

Set a custom session service.

pub fn with_memory_service(self, service: Arc<dyn Memory>) -> Launcher

Set a custom memory service.

pub fn with_compaction(self, config: EventsCompactionConfig) -> Launcher

Enable runner-level context compaction in serve mode.

pub fn with_context_cache( self, config: ContextCacheConfig, cache_capable: Arc<dyn CacheCapable>, ) -> Launcher

Enable automatic prompt cache lifecycle management in serve mode.

pub fn with_security_config(self, config: SecurityConfig) -> Launcher

Set custom server security settings.

pub fn with_request_context_extractor( self, extractor: Arc<dyn RequestContextExtractor>, ) -> Launcher

Set a request context extractor for authenticated deployments.

pub fn with_a2a_base_url(self, base_url: impl Into<String>) -> Launcher

Enable A2A routes when building or serving the HTTP app.

pub fn with_telemetry(self, config: TelemetryConfig) -> Launcher

Configure how serve mode initializes telemetry.

pub fn with_shutdown_grace_period(self, grace_period: Duration) -> Launcher

Set the maximum graceful shutdown window for the web server.

pub fn with_streaming_mode(self, mode: StreamingMode) -> Launcher

Set streaming mode for console mode.

Note: this currently only affects console mode. The server does not yet accept a RunConfig.

pub fn with_thinking_mode(self, mode: ThinkingDisplayMode) -> Launcher

Control how emitted thinking content is rendered in console mode.

pub async fn run(self) -> Result<(), AdkError>

Run the launcher, parsing CLI arguments.

• No arguments or chat: Interactive console
• serve [--port PORT]: Web server with UI

pub async fn run_console_directly(self) -> Result<(), AdkError>

Run in interactive console mode without parsing CLI arguments.

Use this when you already know you want console mode (e.g. from your own CLI parser). run calls this internally.

pub fn build_app(self) -> Result<Router, AdkError>

Build the Axum application without serving it.

This is the production escape hatch for adding custom routes, middleware, metrics, or owning the serve loop yourself.

pub fn build_app_with_a2a( self, base_url: impl Into<String>, ) -> Result<Router, AdkError>

Build the Axum application with A2A routes enabled.

pub async fn run_serve_directly(self, port: u16) -> Result<(), AdkError>

Run web server without parsing CLI arguments.

Use this when you already know you want serve mode (e.g. from your own CLI parser). run calls this internally.