How to Test AI Agents With the TestMu AI Agent Testing CLI
The TestMu AI Agent Testing CLI (agent-testing-cli) tests AI agents from a terminal, with no dashboard clicks. Use it to discover projects and test resources, run Chat and Phone Caller evaluations, retrieve results, and automate tests in a continuous integration and continuous delivery (CI/CD) pipeline.
It runs the same commands on macOS, Linux, and Windows against Chat, Voice, and Phone Caller agents configured in TestMu. You can drive it with direct commands, machine-readable JSON output, or the interactive terminal user interface (TUI), and authenticate with environment variables so suites run without an interactive prompt.
- Run tests from a terminal or IDE: Use the same commands on macOS, Linux, and Windows.
- Test different agent types: Work with Chat, Voice, and Phone Caller agents that are configured in TestMu.
- Choose an operating mode: Use direct commands, machine-readable output, or the interactive terminal user interface (TUI).
- Integrate with CI/CD: Authenticate with environment variables and run suites without an interactive prompt.
# Install
pip install agent-testing-cli
# Authenticate
# Enter your TestMu username and access key when prompted.
agent-testing-cli login
# List your projects
agent-testing-cli projects
How to Set Up Your Credentials
- You need a TestMu AI account. Sign up free if you do not have one.
- Follow Get started with TestMu Agent Testing and open Credentials in the TestMu AI Dashboard. Copy your username and access key.
- Set the credentials as environment variables when you use the CLI in CI/CD or other non-interactive environments.
- macOS or Linux
- Windows PowerShell
export LT_USERNAME="YOUR_USERNAME"
export LT_ACCESS_KEY="YOUR_ACCESS_KEY"
$env:LT_USERNAME = "YOUR_USERNAME"
$env:LT_ACCESS_KEY = "YOUR_ACCESS_KEY"
Which Environments the CLI Supports
The CLI runs in a system terminal or an integrated terminal in an IDE on:
- macOS
- Linux
- Windows
What the CLI Requires
- Python 3.10 or newer
- macOS, Linux, or Windows
- A TestMu account with access to Agent-to-Agent Testing
- At least one TestMu project with an agent and a test suite
How to Choose a Usage Mode
| Mode | Command | Best for |
|---|---|---|
| Interactive TUI | agent-testing-cli tui | Local exploration and suite selection |
| Command mode | agent-testing-cli --project PROJECT_ID run --suite SUITE_ID | Direct execution from a terminal |
| CI/CD mode | agent-testing-cli --project PROJECT_ID --json run ... --yes | Pipelines and scripts |
--yes confirms a test run without an interactive prompt. Use it for CI/CD or after you verify the selected project, suite, workflow, or URL.
Quick Command Reference
| Command | Purpose |
|---|---|
agent-testing-cli login | Validate and save credentials. |
agent-testing-cli auth status | Show the authentication status. |
agent-testing-cli projects | List accessible projects. |
agent-testing-cli --project PROJECT_ID scenarios | List saved scenarios for a project. |
agent-testing-cli --project PROJECT_ID suites | List saved suites for a project. |
agent-testing-cli --project PROJECT_ID endpoint-profiles | List Chat endpoint profiles for a project. |
agent-testing-cli --project PROJECT_ID run --suite SUITE_ID | Start a Phone Caller suite. |
agent-testing-cli results --suite SUITE_ID | Show suite results. |
agent-testing-cli call --id CALL_ID | Show a call result. |
agent-testing-cli tui | Open the interactive TUI. |
agent-testing-cli logout | Delete credentials for the selected profile. |
Run the following command to see all commands:
Verifiedagent-testing-cli --help
Run the following command to see the options for one command:
Verifiedagent-testing-cli COMMAND --help
How to Install the CLI
Install the latest published version from PyPI:
Verifiedpip install agent-testing-cli
Verify the installation:
Verifiedagent-testing-cli --help
How to Authenticate Your Account
Sign In Interactively
Run the login command:
Verifiedagent-testing-cli login
The CLI prompts for your username and access key. The CLI saves credentials in the operating-system credential store.
You can also supply the username and access key as options:
Verifiedagent-testing-cli login --user LT_USERNAME --key LT_ACCESS_KEY
Do not store the access key in source control or shell history.
Authenticate in CI/CD
Set both environment variables in the CI/CD environment:
Verifiedexport LT_USERNAME="your-username"
export LT_ACCESS_KEY="your-access-key"
Store these values in the secret store of your CI/CD platform.
Check Status and Sign Out
Check the authentication status:
Verifiedagent-testing-cli auth status
Delete saved credentials for the selected profile:
Verifiedagent-testing-cli logout
How to Discover Projects and Test Resources
List accessible projects:
Verifiedagent-testing-cli projects
Use a project ID to list its configured resources:
Verifiedagent-testing-cli --project PROJECT_ID scenarios
agent-testing-cli --project PROJECT_ID suites
agent-testing-cli --project PROJECT_ID endpoint-profiles
Add --json when a script must process the output:
agent-testing-cli --json projects
How to Test a Phone Caller Agent
For a Phone Caller run, --project PROJECT_ID identifies the TestMu project and --suite SUITE_ID identifies the suite to run.
Start one saved suite:
Verifiedagent-testing-cli --project PROJECT_ID run \
--suite SUITE_ID
The CLI asks for confirmation because a Phone Caller suite can create real calls. Review the selected project and suite before you confirm the run.
Start a suite without an interactive confirmation:
Verifiedagent-testing-cli --project PROJECT_ID run \
--suite SUITE_ID \
--yes
Start multiple suites:
Verifiedagent-testing-cli --project PROJECT_ID run \
--suite SUITE_ID_1 \
--suite SUITE_ID_2 \
--yes
Wait for the final suite result:
Verifiedagent-testing-cli --project PROJECT_ID run \
--suite SUITE_ID \
--yes \
--wait \
--poll 5 \
--timeout 1800
The --poll and --timeout values are in seconds.
How to Test a Chat Agent
A Chat evaluation requires a project, workflow, and suite:
Verifiedagent-testing-cli --project PROJECT_ID run \
--workflow WORKFLOW_ID \
--suite SUITE_ID \
--yes
Select a specific endpoint profile:
Verifiedagent-testing-cli --project PROJECT_ID run \
--workflow WORKFLOW_ID \
--suite SUITE_ID \
--endpoint-profile ENDPOINT_PROFILE_ID \
--yes
Chat evaluations are asynchronous. Do not use --wait with a Chat evaluation.
How to Run from URLs
Run a suite from a supported TestMu result URL:
Verifiedagent-testing-cli run --url "SUITE URL" --yes
Run multiple suite URLs in one command:
Verifiedagent-testing-cli run --url "SUITE_URL_1" "SUITE_URL_2" "SUITE_URL_3" --yes
The --yes flag confirms every selected run without an interactive prompt. Use it only after you verify the URLs.
How to Use the Interactive TUI
Open the terminal user interface:
Verifiedagent-testing-cli tui
| Key or action | Result |
|---|---|
Tab | Move to the next list or control. |
| Arrow keys | Move through the active list. |
Space | Select or clear a suite. |
Enter or r | Run the selected suites. |
c or y | Copy the visible run command. |
f | Refresh project data. |
q | Close the TUI. |
The TUI shows a copy-ready command for the selected project or suite. The TUI also asks for confirmation before it starts a test run.
What the Global Options Do
The shared --json, --verbose, and --no-tui options can appear before or after a command. Completion options must appear before a command. Use the global form --project PROJECT_ID before the command for project-scoped operations. The existing command-level form remains supported for compatibility.
| Option | Purpose |
|---|---|
--project PROJECT_ID | Select a project for a project-scoped command. |
--json | Write machine-readable JSON. |
--verbose | Write sanitized diagnostics to standard error. |
--no-tui | Disable the interactive TUI. |
--install-completion | Install completion for the current shell. |
--show-completion | Show the completion script for the current shell. |
--help | Show help. |
How to Integrate with CI/CD
For a Chat evaluation, provide the workflow ID as an environment variable and use this command:
Verifiedagent-testing-cli --project "$TESTMU_PROJECT_ID" --json run \
--workflow "$TESTMU_WORKFLOW_ID" \
--suite "$TESTMU_SUITE_ID" \
--yes
Do not add --wait to a Chat evaluation.
What the Exit Codes Mean
| Code | Meaning |
|---|---|
0 | The command completed successfully. |
1 | A completed test failed. |
2 | The command input is invalid. |
3 | Authentication failed or credentials are unavailable. |
4 | An API or TUI dependency error occurred. |
5 | Result polling reached the timeout. |
Use the exit code to fail a CI/CD job when a command does not complete successfully.
How to Troubleshoot Common Problems
Credentials Are Unavailable
What you see: The CLI asks you to log in or reports that credentials are unavailable.
Fix: Run agent-testing-cli login. In CI/CD, set both LT_USERNAME and LT_ACCESS_KEY.
A Project or Suite Is Not Found
What you see: The CLI reports that a project or suite does not exist.
Fix: List the available projects and suites again:
Verifiedagent-testing-cli projects
agent-testing-cli --project PROJECT_ID suites
A Chat Run Requires a Workflow
What you see: The CLI reports Chat run requires --workflow.
Fix: Add the workflow ID:
Verifiedagent-testing-cli --project PROJECT_ID run \
--workflow WORKFLOW_ID \
--suite SUITE_ID \
--yes
JSON Output Requires Confirmation Bypass
What you see: The CLI reports that --yes is required with --json.
Fix: Add --yes to the run command after you verify the project and suite IDs.
Polling Reaches the Timeout
What you see: A Phone Caller run exits with code 5.
Fix: Increase --timeout, or retrieve the result later with the results command.
The Credential Store Is Unavailable
What you see: Interactive login cannot access the operating-system credential store.
Fix: Configure the credential store for the operating system. For CI/CD, use LT_USERNAME and LT_ACCESS_KEY instead.
Related TestMu AI Guides
- Learn how Agent Testing evaluations work.
- Learn how to connect a Chat agent over its API.
- Open the agent-testing-cli package on PyPI.
Chat with Us