cz-cli Installation and Usage Guide

cz-cli is the command-line and AI Agent operational tool for Singdata Lakehouse. It encapsulates Lakehouse capabilities — connection configuration, SQL execution, Schema and table management, Studio task development, task execution inspection, Job diagnostics, and more — into stable CLI commands. Users can operate directly in the terminal, and AI Agents such as Codex, Claude Code, Cursor, Kiro, and Hermes can use natural language to assist with data warehouse development and operations.

With cz-cli, users can delegate tasks like "create a test data warehouse", "check why today's tasks failed", "backfill data for a time range", "query table schema and sample data" to cz-cli for execution; cz-cli performs the actual operations and returns structured results.

Target Users

• Data development, data platform, operations, and analytics teams who need to manage Singdata Lakehouse via the command line.
• Users who want to connect AI Agents to Singdata, enabling agents to execute queries, create tasks, diagnose runtime issues, and generate operational solutions.
• Teams needing to automate Singdata operations locally, in CI/CD, corporate bots, or operational dashboards.

Core Capabilities

Capability	Description	Common Commands
Connection Management	Create and switch connection profiles for different environments (e.g., production, testing, UAT)	cz-cli profile create, cz-cli profile list, cz-cli profile use
SQL Queries	Execute SELECT, DDL, DML; supports sync wait and async Job	cz-cli sql, cz-cli job status, cz-cli job result
Schema & Table Management	View, create, describe, preview, and gather statistics on Schemas and tables	cz-cli schema, cz-cli table
Studio Tasks	Create SQL, offline integration, real-time sync, and other tasks; configure scheduling, publish online, and manually execute	cz-cli task
Run Inspection	View task execution records, logs, dependencies, statistics; retry on failure and backfill data	cz-cli runs, cz-cli attempts
Performance Diagnostics	View SQL Job status, results, and execution profile	cz-cli job, cz-cli sql --job-profile
AI Agent Integration	Enable Agents to invoke Singdata capabilities using natural language	cz-cli agent run
Data Source Management	Manage external data sources, preparing for sync and import tasks	cz-cli datasource

Prerequisites

Before installation, confirm you have the following information. Refer to the documentation if unsure:

Item	Description	Example
Service Endpoint	Singdata API service address	cn-shanghai-alicloud.api.singdata.com
Instance Name	Singdata instance name or ID	demo_instance
Workspace	Workspace name	analytics_prod
Username & Password	Account credentials for connecting to Singdata	data_user
Default Schema	Schema to use after login	public
Default Compute Group	Virtual Cluster used for SQL execution or tasks	DEFAULT

If using a Personal Access Token (PAT), you can replace username and password with the PAT.

Installing cz-cli

macOS / Linux

Use the one-line install script:

curl -fsSL https://cz-cli.ai/install.sh | bash

After installation, reload your Shell configuration:

zsh Users

source ~/.zshrc

bash Users

source ~/.bashrc

Windows

The install script must run in a Bash environment. Run it in WSL (Windows Subsystem for Linux) or Git Bash:

curl -fsSL https://cz-cli.ai/install.sh | bash

Verify Installation

cz-cli --version

Seeing a version number indicates a successful installation.

You can also check the help:

cz-cli --help

Configuring Connection Profiles

A Profile is a local configuration that stores connection information for cz-cli. We recommend creating a separate profile for each environment, e.g., prod, uat, dev.

Create a Profile Using cz-cli setup

For example:

cz-cli setup

Then follow the guided prompts to register a new Singdata account, or enter your login address, credential string (generated after login), or JDBC connection string from an existing account.

Create with CLI Credential

cz-cli setup --credential <your cli Connection String>

How to Obtain a PAT

1) Log in to your Singdata account, and select LakehouseMCP from the left menu.

2) If you have not created a PAT, click Create PAT to generate one.

3) After creation, switch to the CLI tab and copy your CLI connection string.

4) If you already have a PAT, paste it into the popup window and click Generate Connection String.

Once you have copied the CLI connection string, return to the cz-cli window, insert the connection string into the following command, and execute:

cz-cli setup --credential <your cli Connection String>

Once execution completes, the profile configuration is done.

Create from JDBC Connection String

If you already have a JDBC connection string, you can create a profile directly:

cz-cli profile create prod --jdbc "jdbc:clickzetta://<instance_name>.<service_endpoint>/<workspace_name>?username=<username>&password=<password>&schema=public&virtualCluster=DEFAULT"

Viewing and Switching Profiles

View profiles configured on this machine

cz-cli profile list

Set the default profile

cz-cli profile use prod

Temporarily use a specific profile for a command

cz-cli -p uat status

Verify connection

cz-cli -p prod status

If connected is true in the result, the connection is successful.

First-Time Usage

View the current workspace

cz-cli -p prod workspace list

View Schemas

cz-cli -p prod schema list

Execute a read-only query

cz-cli sql defaults to synchronous execution (--sync), waiting and returning results directly; for large or long-running queries, add --async to return only a job_id and fetch the results later.

cz-cli -p prod sql "SELECT current_timestamp()" --sync

You can also pass SQL using -e:

cz-cli -p prod sql -e "SELECT * FROM public.your_table LIMIT 10" --sync

Execute write operations

To prevent accidental operations, write operations such as INSERT, UPDATE, DELETE, CREATE, DROP require explicitly adding --write.

cz-cli -p prod sql --write --sync -e "CREATE TABLE IF NOT EXISTS public.demo_orders (id INT, amount DECIMAL(18,2))"

View table structure and sample data

cz-cli -p prod table describe public.demo_orders cz-cli -p prod table preview public.demo_orders

Studio Task Development and Operations

cz-cli can operate tasks in Singdata Studio, suitable for data development and daily operations.

Create a SQL Task

A task must be created inside a folder using --folder. First list the existing folders (or create one with cz-cli -p prod task create-folder <name>):

cz-cli -p prod task folder-tree

Then create the task under a specific folder:

cz-cli -p prod task create daily_order_summary --type SQL --description "Daily order summary" --folder <folder_name_or_id>

Save Task SQL

cz-cli -p prod task save-content daily_order_summary --content "INSERT INTO public.order_summary SELECT current_date(), COUNT(*) FROM public.orders"

Configure Scheduling

The following example schedules execution daily at 02:00:

cz-cli -p prod task save-cron daily_order_summary --cron "0 0 2 * * ? *"

Deploy Task

cz-cli -p prod task deploy daily_order_summary

Manually Execute Task

cz-cli -p prod task execute daily_order_summary --max-wait-seconds 300

Viewing Execution Records and Logs

View Recent Runs

cz-cli -p prod runs list --task daily_order_summary --limit 5

View Run Details

cz-cli -p prod runs detail <run_id>

View Run Logs

cz-cli -p prod runs logs <run_id>

Wait for a Run to Complete

cz-cli -p prod runs wait <run_id>

Using with AI Agents

A key value of cz-cli is giving AI Agents a controllable, auditable entry point for operating Singdata. Users can have Agents invoke cz-cli via command line within tools such as Codex, Claude Code, Cursor, Kiro, and Hermes.

Recommended Agent Prompt

Share the following prompt with your AI Agent:

You can use cz-cli to operate Singdata Lakehouse. First run cz-cli status to verify the connection. For write operations, task deployment, backfill, deletion, decommissioning, and other high-risk actions, you must present the execution plan first and wait for my confirmation. Queries and inspections can be executed directly. Present results as concise tables or bullet-point summaries.

Common Natural Language Requests

In the examples below, <profile>, <task_name>, and <job_id> should be replaced with the user's own connection profile, task name, and Job ID. It is recommended to have the Agent first execute cz-cli -p <profile> status to confirm the connection is available.

Using the <profile> environment, first confirm the cz-cli connection status, then list the schemas in the current workspace, and then list the first 20 tables under the public schema. Using the <profile> test environment, help me create an order details table in the demo schema, insert a few rows of test data, and verify that the data can be queried. Before executing write operations, present the plan and wait for my confirmation. Using the <profile> environment, check whether task <task_name> has failed runs today. If so, review the run details and logs, and provide the failure cause and suggested fixes. Using the <profile> environment, analyze the execution of SQL job <job_id>. Check the job status, results, and execution profile to determine if there are performance bottlenecks.

Executing via cz-cli's Built-in Agent Entry Point

If the current environment already has the LLM parameters configured for the cz-cli Agent, you can also run directly:

cz-cli -p <profile> agent run "Help me check today's failed scheduled tasks and categorize them by failure reason"

Using in Enterprise Bot Scenarios

If you use enterprise bots such as Hermes to host AI Agents, we recommend installing cz-cli in the bot execution environment and adopting the following strategies:

• Grant the bot only the necessary Singdata permissions; avoid using high-privilege administrator accounts.
• Enable manual confirmation for write, delete, deploy, decommission, backfill, and similar operations.
• Implement whitelisting or approval controls for users who can access the bot.
• Store credentials such as profiles, PATs, and passwords in controlled environment variables or local configuration — never in public documentation, chat history, or code repositories.

Output Formats and Automation

cz-cli outputs JSON by default, making it easy for AI Agents and scripts to parse. Other formats can also be specified:

Table format, suitable for human reading

cz-cli -p prod --format table status

CSV format, suitable for export

cz-cli -p prod --format csv sql "SELECT * FROM public.orders LIMIT 100" --sync

Extract a single field

cz-cli -p prod --field data.connected status

In CI/CD or automation scripts, we recommend:

• Using fixed profile names, e.g., prod-readonly, uat-admin.
• Using --sync and --timeout for query commands to control wait time.
• Explicitly adding --write for write operations, with approval or manual confirmation retained in the workflow.
• Prioritizing JSON output to avoid parsing natural language text.

Upgrading

Check the current version:

cz-cli --version

Upgrade to the latest version:

cz-cli update

You can also re-run the install script to upgrade to the latest version:

curl -fsSL https://cz-cli.ai/install.sh | bash

FAQ

Q: After installation, I get "cz-cli: command not found"?

Usually the PATH has not taken effect. Reload your Shell configuration:

source ~/.zshrc

Or manually add to PATH:

echo 'export PATH="$HOME/.clickzetta/bin:$PATH"' >> ~/.zshrc source ~/.zshrc

Q: The status command shows connected: false?

Check the following in order:

cz-cli profile list cz-cli -p <profile_name> status

Verify that the service, instance, workspace, username, and password or PAT in the profile are correct. If your Lakehouse instance has network policies configured, confirm that your machine's network can access the Lakehouse service.

Q: How do I modify fields in a profile?

The profile configuration file is located at the .clickzetta/profiles.toml path under your current user directory.

You can modify it using the profile update command:

cz-cli profile update prod workspace <new_workspace_name> cz-cli profile update prod password '<new_password>' cz-cli profile update prod schema public cz-cli profile update prod vcluster DEFAULT

Q: Queries succeed, but task-related commands fail?

Commands like task, runs, and attempts depend on Studio capabilities. Verify that the environment corresponding to the current profile has Studio task capabilities enabled, and that the account has permissions to view or manage tasks.

Q: CREATE TABLE or INSERT is rejected?

Write operations require adding --write:

cz-cli -p prod sql --write --sync -e "INSERT INTO public.demo_orders VALUES (1, 99.9)"

Q: SQL returns a job_id but no data directly?

This is because execution defaults to asynchronous mode. To get query results returned directly, add --sync:

cz-cli -p prod sql "SELECT * FROM public.demo_orders LIMIT 10" --sync

If you already have a job_id, you can continue with:

cz-cli -p prod job status <job_id> cz-cli -p prod job result <job_id>

Q: How can I reduce the risk of misoperations?

• Queries, inspections, and diagnostics can be executed by the Agent directly.
• For write, delete, task deploy, task decommission, backfill, and retry-on-failure actions, we recommend requiring the Agent to first present a plan and wait for manual confirmation.
• Configure a separate low-privilege account or read-only profile for the Agent.
• Use explicit profile names for production environments, e.g., prod-readonly, prod-operator.

Recommended Getting Started Path

1. Install cz-cli and verify that cz-cli --version works.
2. Create a profile and verify the connection with cz-cli -p <profile> status.
3. Use schema list, table list, and sql --sync to complete a read-only query.
4. In a test environment, try creating a table or inserting data to become familiar with the --write protection mechanism.
5. Review task --help and runs --help to understand task development and operations commands.
6. Hand cz-cli over to your AI Agent, with an agreement that high-risk operations require manual confirmation.

For concepts such as account name (account_name) and service name (instance_name) and how to find them, see: https://www.singdata.com/documents/key-concepts

---

Related Documentation

• SQL Execution and Data Exploration — Complete command reference for sql, schema, table, job, workspace
• Studio Task Development and Operations — Task creation, scheduling, runs operations, backfill, task flow
• AI Agent Integration — Agent LLM configuration, natural language operations, enterprise bot scenarios
• External Data Source Management — Data source browsing, connectivity testing, sample data preview