All Products
Search
Document Center

Server Migration Center:SMC AI Agent Edition Client (Public Preview)

Last Updated:Aug 10, 2026

The Server Migration Center (SMC) AI Agent edition client (SMCClaw) is an intelligent migration assistant. It migrates servers to the cloud through conversation. The traditional SMC client uses a command line. SMCClaw lets you chat instead: import migration sources, create and manage migration tasks, monitor status, and troubleshoot issues. It also supports scheduled tasks.

Client overview

Features

  • AI-driven migration: Import migration sources, create tasks, query status, and troubleshoot issues through natural-language conversation.

  • SMC API operations: Built-in SMC migration skills call SMC APIs to manage migration sources and tasks.

  • Scheduled tasks: Track and monitor the progress of specified migration tasks.

  • Web console: A dashboard for AI conversation, session management, channel configuration, scheduled tasks, model configuration, and log viewing.

  • Command-line TUI: A terminal interface for chatting directly with the AI assistant.

  • Command approval: An allowlist-based approval mechanism that gates risky commands.

Comparison with the traditional SMC client

Item

Traditional SMC client

SMC AI Agent edition client

Interaction

Command line. You enter parameters manually.

Natural-language conversation. The AI interprets your intent and acts.

Import migration source

Run an activation-code command manually.

Import with a single conversational instruction.

Manage migration tasks

Use the SMC console or the command line.

Manage through conversation,.

Status monitoring

Log in to the console manually.

Scheduled tasks query automatically.

Troubleshooting

Investigate in the console or open a ticket.

The AI analyzes and fixes migration errors automatically.

Skill level

Requires knowledge of SMC migration steps and parameters.

Describe your needs in plain language. The AI guides you.

Scenarios

  • Server migration: Migrate a local or remote server to the cloud.

  • Migration progress monitoring: Set a scheduled task to monitor the progress of a specific migration task.

  • Migration troubleshooting: Ask about migration issues and troubleshooting.

Install and configure the client

Step 1: Prepare the environment

  1. Prepare a server.

    • Linux: Public network access to reach Alibaba Cloud SMC and your chosen large language model (LLM) API. x86_64 architecture, kernel 5.10 or later, Python 3.9 or later. We recommend Alibaba Cloud Linux 3, Ubuntu 22.04, Rocky Linux 9, or later.

    • Windows: Public network access to reach Alibaba Cloud SMC and your chosen LLM API. x86_64 architecture, Windows Server 2012 R2 or later. We recommend Windows Server 2016, Windows 10, or later.

  2. Prepare an Alibaba Cloud account with SMC enabled. SMC itself is free, but the intermediate instances, cloud disks, and network traffic used during migration may incur charges.

  3. Prepare an LLM API key. We recommend the Alibaba Cloud (Qwen Cloud) Token Plan or Coding Plan.

Step 2: Download and install

Linux

Installation and initial setup require administrator privileges. Run the following commands as the root user to download and install the client.

You can also download the package directly from the Linux x64 installer and upload it to the server.
  1. Download the AI Agent edition client.

    curl -fL -o smcclaw-2026.7.1-linux-x64.tar.gz \
      "https://p2v-tools.oss-cn-hangzhou.aliyuncs.com/smc/smcclaw/2026.7.1/smcclaw-2026.7.1-linux-x64.tar.gz"
  2. Extract and install the client.

    tar -xzf smcclaw-2026.7.1-linux-x64.tar.gz
    bash smcclaw-2026.7.1-linux-x64/install.sh
  3. Check the client version.

    smcclaw --version

Windows

Installation and initial setup require administrator privileges. Run the following commands as an administrator in PowerShell to download and install the client.

You can also download the package directly from the Windows x64 installer and upload it to the server.
  1. Download the AI Agent edition client.

    Invoke-WebRequest `
      -Uri "https://p2v-tools.oss-cn-hangzhou.aliyuncs.com/smc/smcclaw/2026.7.1/smcclaw-2026.7.1-win32-x64.zip" `
      -OutFile "smcclaw-2026.7.1-win32-x64.zip"
  2. Extract and install the client.

    Expand-Archive "smcclaw-2026.7.1-win32-x64.zip" -DestinationPath "C:\smcclaw"
    Set-Location "C:\smcclaw\smcclaw-2026.7.1-win32-x64\bin"
  3. Check the client version.

    .\smcclaw.cmd --version
Note: To call smcclaw globally, add its path to the PATH environment variable. Otherwise, go to the bin directory and run .\smcclaw.cmd. This topic assumes that PATH is already configured.

Step 3: Configure the client

  1. Configure the LLM.

    One-click setup (recommended)

    The following examples use the Coding Plan (Global) with the model qwen/qwen3.7-plus. For more plans and models, see More model configurations.

    Use one plan only. Do not pass multiple model keys in a single setup command.

    Parameter

    Description

    <AccessToken ID>

    The AccessToken ID of the SMC activation code. To obtain it, see obtain an activation code.

    <AccessToken Code>

    The AccessToken Code of the SMC activation code. To obtain it, see obtain an activation code.

    <API_KEY>

    For the Global API key, see the First API call.

    --model

    The model to use, in the format provider/model. For example, --model "qwen/qwen3.7-plus".

    smcclaw setup --modelstudio-api-key "<API_KEY>" --model "qwen/qwen3.7-plus" --accesstokenid "<AccessToken ID>" --accesstokencode "<AccessToken Code>"
    Important

    The activation code is a sensitive credential. Entering it directly in a command may save it to your shell history. Manage credentials according to your security requirements.

    Interactive setup with onboard

    The following examples use the Coding Plan (Global) with the model qwen/qwen3.7-plus.
    1. Start the interactive setup.

      To obtain <AccessToken ID> and <AccessToken Code>, see obtain an activation code.

      smcclaw onboard --accesstokenid "<AccessToken ID>" --accesstokencode "<AccessToken Code>"
      Important

      The activation code is a sensitive credential. Entering it directly in a command may save it to your shell history. Manage credentials according to your security requirements.

    2. Follow the prompts:

      • Setup mode: OneClick

      • Model/auth provider: Qwen Cloud (if not listed, select More.)

      • Qwen Cloud auth method: Coding Plan API Key for Global/Intl

      • API key: Enter your Coding Plan API key.

      • Default model: Select a model, such as qwen/qwen3.7-plus.

  2. Run smcclaw gateway status. A running status means the Gateway service has started.

Step 4: Start a conversation

The Gateway listens on 127.0.0.1:18789 by default and offers three conversation entry points.

Dashboard (recommended)

  1. Get the dashboard address.

    smcclaw dashboard --no-open --print-auth-url

    The Authenticated Dashboard URL in the output is the dashboard address.

  2. Open the dashboard in a browser.

    • Linux: If the server has no browser, run the following command on a machine that has a browser and can reach SMCClaw over SSH to create a tunnel. Keep the command running, then open the dashboard URL in your browser.

      ssh -N -L 18789:127.0.0.1:18789 <SMCClaw-account>@<SMCClaw-IP-or-hostname>
    • Windows: Open the dashboard URL directly in the server's browser.

TUI

Run the following command to start the TUI on the local machine:

smcclaw tui

Common TUI commands

Command

Purpose

/help

View help.

/status

View status.

/new

Start a session.

/model <model>

Switch the model.

/stop

Stop the current task.

/exit

Exit the TUI.

The TUI cannot receive scheduled-task notifications. To monitor migration progress automatically, use the dashboard or an IM channel.

Migrate a server for the first time

SMCClaw supports three server migration scenarios:

  • Direct local migration: Migrate the server that runs SMCClaw.

  • Remote SSH migration: Migrate a remote server over SSH.

  • Remote assisted migration: SMCClaw provides an import script that you run manually before migration.

Example instructions:

Migrate this server to the Alibaba Cloud Beijing region.
I need to migrate another cloud host over SSH.
I have an on-premises Windows server with no remote connection.

Migration workflow

  1. Identify the migration intent: The AI interprets your conversation and confirms the migration target.

    Migration source registration, migration drills, and data synchronization consume CPU, disk, and network resources. For disk shrinking during migration, assess the file system, partition layout, actual data volume, and drill results together. In production, run migration during off-peak hours and back up your data first.
  2. Configure the activation code: On the first migration, if no SMC activation code is saved locally, the client asks for the AccessToken ID and AccessToken Code. Obtain the activation code and provide it through the conversation.

  3. Register or import the migration source: After the AI collects source information, it imports the migration source and returns its basic details.

  4. Analyze the migration and estimate costs: The AI analyzes the migration and estimates costs based on the source information and cloud resource prices.

    The estimate is for reference only. Actual charges are based on your Alibaba Cloud bill.
  5. Create and start the migration task: After you confirm the costs, the AI creates and starts the migration task and sets a scheduled task to monitor progress automatically. The task runs a migration drill by default: it first tests migration feasibility and produces a drill report, then proceeds to formal migration automatically after the drill passes.

  6. Monitor migration progress: During migration, the scheduled task periodically queries progress in the current session and reports it.

  7. Complete the migration: When the migration task finishes, the AI cancels the progress-monitoring scheduled task, returns a migration summary, and waits for your next instruction.

More client configurations

Scheduled tasks

When migration starts, SMCClaw creates a scheduled task to monitor the specified migration task by default. You can also create custom scheduled tasks through conversation:

Start migration task j-bp1xxx tonight at 22:00 and notify me when it finishes.
Check the status of all migration sources daily at 9:00 and alert me on anomalies.

You can view the scheduled tasks you created in the dashboard console.

Common management commands

# List all scheduled tasks
smcclaw cron list

# View details of a task
smcclaw cron get <jobId>

# Run a task once immediately
smcclaw cron run <jobId>

# View the execution history of a task
smcclaw cron runs --id <jobId>

# Enable a disabled task
smcclaw cron enable <jobId>

# Disable a task (pause without deleting)
smcclaw cron disable <jobId>

# Delete a task permanently
smcclaw cron remove <jobId>

Common Cron expressions

Expression

Description

0 9 * * *

Every day at 9:00

0 9 * * 1-5

Weekdays at 9:00

0 */2 * * *

Every 2 hours

30 22 * * *

Every day at 22:30

Important

Scheduled tasks depend on a running Gateway. Run smcclaw cron runs --id <jobId> to view the execution history.

Security policies

  • Gateway binding: The Gateway binds only to the local loopback address by default. For remote access, use an SSH tunnel first.

  • Gateway token: The Gateway token and any dashboard URL that contains the token are sensitive credentials. Do not share them or keep them in unredacted screenshots.

  • Activation code: The SMC activation code is a sensitive credential. Provide it only in a local TUI or a private dashboard session, and clean up documents, screenshots, and logs regularly.

  • Command source: Do not submit commands from unknown sources to the AI. Verify the full command and target path before you run a destructive operation.

  • Execution policy: The default policy is allowlist + on-miss: any command outside the allowlist requires approval. We recommend keeping it enabled.

To view the security configuration, run:

smcclaw config get tools.exec

Start and stop the service

# Start the service
smcclaw gateway start

# Stop the service
smcclaw gateway stop

# Restart the service
smcclaw gateway restart

# Check the service status
smcclaw gateway status

Uninstall the client

  • Linux: Run the following commands to uninstall the client.

    # Remove the Gateway service first
    smcclaw gateway uninstall
    
    # Then delete the program files
    bash /opt/smcclaw/install.sh --uninstall
    Important

    If you used a custom directory during installation, specify the same --prefix directory when you uninstall. Otherwise, the cleanup fails. For example: bash /usr/local/smcclaw/install.sh --prefix /usr/local/smcclaw --uninstall

  • Windows: Run smcclaw gateway uninstall to remove the Gateway service, then delete the extracted SMCClaw directory.

Clean up configuration and data

Run the following commands to back up and delete the configuration or working directory:

smcclaw backup create
smcclaw uninstall --dry-run
smcclaw uninstall

smcclaw uninstall removes the Gateway service and the selected local data. It does not delete the CLI program files. To delete the CLI program files, see uninstall the client.

Status query and diagnostics

# View the SMCClaw runtime status and configuration
smcclaw status

# Run self-diagnostics to check the runtime environment
smcclaw doctor

# View the SMCClaw runtime logs
smcclaw logs --follow

# View the current AI model configuration status
smcclaw models status

# View the message channel connection status
smcclaw channels status

FAQ

How do I obtain an activation code?

  1. Go to the Server Migration page in the SMC console.

  2. On the Server Migration page, click Import Source Servers.

  3. On the Import Source Servers page, click Use Client to Import Server.

  4. On the Import Migration Source page, select Client Import.

  5. In the Description, select the operating system type, then click Generate.

  6. Copy the generated Access Token ID and Access Token Code.

How do I customize the installation directory on Linux?

By default, the installation script:

  • Installs to /opt/smcclaw.

  • Creates a symbolic link at /usr/local/bin/smcclaw.

  • Attempts to install the Python dependencies for the SMC migration skills.

To customize the installation directory, run:

bash smcclaw-2026.7.1-linux-x64/install.sh --prefix /usr/local/smcclaw

More model configurations

Recommended: Alibaba Cloud (Qwen Cloud)

The Qwen provider is built in. The first setup does not prompt you to download a plugin from npm. If you upgraded from an older version that downloaded the Qwen plugin automatically, run smcclaw doctor --fix first to clear the old plugin installation record. Choose one command based on your plan and site:

Plan

Site

One-click setup command

Token Plan (recommended)

China

smcclaw setup --qwen-token-plan-api-key-cn "<API_KEY>" --model "qwen/qwen3.7-plus" --accesstokenid "<AccessToken ID>" --accesstokencode "<AccessToken Code>"

Coding Plan

China

smcclaw setup --modelstudio-api-key-cn "<API_KEY>" --model "qwen/qwen3.7-plus" --accesstokenid "<AccessToken ID>" --accesstokencode "<AccessToken Code>"

Coding Plan

Global

smcclaw setup --modelstudio-api-key "<API_KEY>" --model "qwen/qwen3.7-plus" --accesstokenid "<AccessToken ID>" --accesstokencode "<AccessToken Code>"

Standard API

China

smcclaw setup --modelstudio-standard-api-key-cn "<API_KEY>" --model "qwen/qwen3.7-plus" --accesstokenid "<AccessToken ID>" --accesstokencode "<AccessToken Code>"

Standard API

Global

smcclaw setup --modelstudio-standard-api-key "<API_KEY>" --model "qwen/qwen3.7-plus" --accesstokenid "<AccessToken ID>" --accesstokencode "<AccessToken Code>"

Important
  • Choose one command only. Do not pass multiple model keys in a single setup command.

  • The Token Plan uses the OpenAI-compatible endpoint https://token-plan.cn-beijing.maas.aliyuncs.com/compatible-mode/v1. You can get the API key from the Model Studio console.

  • The --model parameter specifies the model in the format provider/model. For example, --model "qwen/qwen3.7-plus".

Other mainstream models

Model service

One-click setup command

OpenAI

smcclaw setup --openai-api-key "<API_KEY>"

Anthropic Claude

smcclaw setup --anthropic-api-key "<API_KEY>"

Google Gemini

smcclaw setup --gemini-api-key "<API_KEY>"

DeepSeek

smcclaw setup --deepseek-api-key "<API_KEY>"

Moonshot/Kimi (China)

smcclaw setup --auth-choice moonshot-api-key-cn --moonshot-api-key "<API_KEY>"

Moonshot/Kimi (Global)

smcclaw setup --moonshot-api-key "<API_KEY>"

Zhipu Z.AI

smcclaw setup --zai-api-key "<API_KEY>"

Volcengine

smcclaw setup --volcengine-api-key "<API_KEY>"

MiniMax (China)

smcclaw setup --auth-choice minimax-cn-api --minimax-api-key "<API_KEY>"

Mistral

smcclaw setup --mistral-api-key "<API_KEY>"

Custom OpenAI-compatible endpoint

smcclaw setup --custom-base-url "<BASE_URL>" --custom-model-id "<MODEL_ID>" --custom-api-key "<API_KEY>"

After configuration, check the status:

smcclaw models status
smcclaw doctor
Important

Entering an API key directly in a command may save it to your shell history. Manage credentials according to your security requirements.

How do I check model usage?

Open the dashboard usage page at http://127.0.0.1:18789/usage to view the API token usage and cost of each model.

The dashboard shows a connection failure. What do I do?

If you see "The browser cannot complete the Gateway connection. Check the target and transport before you retry the credentials," check whether the Gateway token is correct.

The AI reply fails with "Agent failed before reply." What do I do?

Verify that the model configuration is correct. Check whether you used the wrong model API key or provider/model name.

The Gateway connection fails. What do I do?

  1. Confirm the service status.

    Run smcclaw gateway status to check whether the Gateway is running.

  2. Check the listening port.

    The Gateway listens on port 18789 by default. Check whether the port is in use, blocked by a firewall, or not being listened on by the Gateway.

  3. Run environment self-diagnostics.

    Run smcclaw doctor to check the environment configuration for issues.

  4. View the runtime logs to find the cause of the connection failure. Run smcclaw logs --follow.

The model call fails. What do I do?

  1. Check whether the model status is normal.

    smcclaw models status
  2. Run environment self-diagnostics.

    smcclaw doctor
  3. View the runtime logs.

    smcclaw logs --follow
  4. Verify the API key.

    Check whether the AccessKey is valid, expired, or disabled.

A scheduled task did not run. How do I troubleshoot it?

  1. View the task list.

    smcclaw cron list

    Confirm that the task was created, is enabled, and has the correct schedule.

  2. View the details of a single task.

    smcclaw cron get <jobId>

    Check the schedule rule, last run time, next scheduled run time, and configuration.

  3. Confirm the Gateway status.

    smcclaw gateway status

    Scheduled tasks are scheduled by the Gateway. If the Gateway is not running or is abnormal, no scheduled task triggers.

  4. Check the server time zone.

    Task run times depend on the server time zone. If the time zone differs from what you expect, tasks run at the wrong time and appear not to run.

  5. View the execution history.

    smcclaw cron runs --id <jobId>

How do I add smcclaw to the system PATH on Windows?

Add the smcclaw executable path to the PATH environment variable to call it globally from the command line. The following

Example uses Windows Server 2022.
  1. Right-click the image.png and select System.

  2. Search for and open View advanced system settings in the left-side navigation pane.

  3. On the Advanced tab of System Properties, click Environment Variables.

  4. Under System variables, find the Path variable. Select it, click Edit, click New, and add C:\smcclaw\smcclaw-2026.7.1-win32-x64\bin.

  5. Test whether the environment variable is configured correctly.

    • Right-click the image.png and select Run. Enter PowerShell to open a new PowerShell window.

    • Run smcclaw --version. If the tool version appears, the environment variable is configured correctly.

Related topics