Use the Command Line¶

ADK provides an interactive terminal interface for testing your agents. This is useful for quick testing, scripted interactions, and CI/CD pipelines.

Run an agent¶

Use the following command to run your agent in the ADK command line interface:

adk run my_agent

npx @google/adk-devtools run agent.ts

import (
    "google.golang.org/adk/v2/cmd/launcher"
    "google.golang.org/adk/v2/cmd/launcher/full"
)

func main() {
    // ... build your agent and config ...
    l := full.NewLauncher()
    if err := l.Execute(ctx, config, os.Args[1:]); err != nil {
        log.Fatalf("Run failed: %v\n\n%s", err, l.CommandLineSyntax())
    }
}

go run agent.go           # console is the default sublauncher
go run agent.go console   # or explicitly name the console subcommand

mvn compile exec:java -Dexec.mainClass="com.example.agent.AgentCliRunner"

This starts an interactive session where you can type queries and see agent responses directly in your terminal.

Running agent my_agent, type exit to exit.
[user]: What's the weather in New York?
[my_agent]: The weather in New York is sunny with a temperature of 25°C.
[user]: exit

Running agent my_agent, type exit to exit.
[user]: What's the weather in New York?
[my_agent]: The weather in New York is sunny with a temperature of 25°C.
[user]: exit

User -> What's the weather in New York?

Agent -> The weather in New York is sunny with a temperature of 25°C.

User ->

Running agent my_agent, type exit to exit.
[user]: What's the weather in New York?
[my_agent]: The weather in New York is sunny with a temperature of 25°C.
[user]: exit

Session options¶

The adk run command includes options for saving, resuming, and replaying sessions.

Save sessions¶

To save the session when you exit:

adk run --save_session path/to/my_agent

You'll be prompted to enter a session ID, and the session will be saved to path/to/my_agent/<session_id>.session.json.

You can also specify the session ID upfront:

adk run --save_session --session_id my_session path/to/my_agent

Resume sessions¶

To continue a previously saved session:

adk run --resume path/to/my_agent/my_session.session.json path/to/my_agent

This loads the previous session state and event history, displays it, and allows you to continue the conversation.

Replay sessions¶

To replay a session file without interactive input:

adk run --replay path/to/input.json path/to/my_agent

The input file should contain initial state and queries:

{
  "state": {"key": "value"},
  "queries": ["What is 2 + 2?", "What is the capital of France?"]
}

Storage options¶

Example with storage options¶

adk run --session_service_uri "sqlite:///my_sessions.db" path/to/my_agent

All options¶

go run agent.go console -streaming_mode none

go run agent.go -streaming_mode sse

Usage telemetry¶

The ADK CLI collects anonymous usage telemetry to understand feature adoption, guide development priorities, and improve tool performance. Data collection is OFF by default until you explicitly choose to enable it.

Your telemetry preference is stored locally on your machine in ~/.adk/config.json. You can manage telemetry data collection at any time through the terminal:

• Enable: adk telemetry enable
• Disable: adk telemetry disable
• Check status: adk telemetry status

You can also manually deactivate telemetry data collection at any time by opening ~/.adk/config.json and setting the telemetry attribute to false:

{
  "telemetry": false
}

What data is collected

• Environment Properties: Operating system information, runtime language and version, and installed ADK CLI version.
• Command Execution Events: Generic command and subcommand names, flags passed, execution duration, exit codes, and exception types if an error occurs. We also log a sequence number and an ephemeral session ID that is discarded after command execution.

What data is not collected

The CLI does not collect sensitive, private, or personal data, specifically:

• Arguments or parameter values passed to commands or flags, such as agent names, prompt strings, file paths.
• User credentials, usernames, API keys, OAuth tokens, or secrets.
• Google Cloud Project IDs or Cloud Account details.
• Source code files, file contents, or directory paths.
• Personally Identifiable Information (PII).