AgentLoop:Onboarding AI Coding Agents

Overview of LoongSuite Pilot

LoongSuite Pilot is a lightweight collector that runs on developer machines. It automatically discovers installed AI Coding Agents and deploys the required hooks or plugins. LoongSuite Pilot normalizes raw activity data from different agents into a unified event format and reports the data to the log or trace backend used by AgentLoop.

Scenarios

When your team uses AI Coding Agents such as Claude Code, Codex, Cursor, or Qoder for daily development, you can use LoongSuite Pilot to collect client-side activity data for the following scenarios:

• View which AI Coding Agents are being used across the team.

• Analyze sessions, model calls, tool calls, and error events.

• Track token usage broken down by agent.

• Investigate agent runtime behavior through log queries, trace links, or custom interfaces.

• Filter sensitive content such as prompts, completions, and tool parameters before data is reported.

Supported AI Coding Agents

The following tables list the agents and data collection capabilities supported by LoongSuite Pilot on Linux / macOS and Windows, respectively.

Linux / macOS Support

Windows Support

Prerequisites

Before you begin, make sure that the following general requirements are met:

• The target AI Coding Agent is installed and has been used at least once.

• The AI Coding Agent onboarding resources have been initialized in the AgentLoop console, and you have obtained the auto-generated installation command.

macOS / Linux

Make sure that the following requirements are also met on your machine:

• You usually do not need to pre-install Node.js or npm. The installer downloads and pins a managed Node.js runtime and prebuilt dependencies.

• curl or wget is installed.

• You can access the OSS download URLs in the installation command.

Windows

Make sure that the following requirements are also met on your machine:

• Windows 10 or later.

• PowerShell 5.1 or later.

• You usually do not need to pre-install Node.js or npm, and you can access the OSS download URLs in the installation command.

Fallback: If the managed Node.js runtime is unavailable because the platform is not supported or the download fails, the installer falls back to the system Node.js. In this case, Node.js 18 or later and npm must be installed.

Known platform limits: Windows ARM64 and Linux distributions that use musl (for example, Alpine) currently have no matching managed Node.js artifact. On these platforms, prepare system Node.js 18 or later and npm in advance.

Install LoongSuite Pilot

For AgentLoop users, we recommend that you copy the installation command auto-generated in the AgentLoop console. The resource names and connection parameters in the command (except true and false) are generated automatically during onboarding resource initialization.

Obtain the CMS and SLS parameter values from the AgentLoop console.

macOS / Linux

# Run the installation command. Replace all parameter values with the ones obtained above.
curl -fsSL https://aliyun-observability-release-cn-shanghai.oss-cn-shanghai.aliyuncs.com/loongsuite-pilot/installer.sh -o /tmp/loongsuite-pilot-installer.sh && bash /tmp/loongsuite-pilot-installer.sh install \
  --collect-log "true" \
  --collect-trace "true" \
  --sls-project "your_SLS_Project_name" \
  --sls-logstore "your_SLS_Logstore_name" \
  --sls-endpoint "your_SLS_endpoint" \
  --cms-license-key "your_Trace_License_Key" \
  --cms-endpoint "your_Trace_endpoint" \
  --cms-workspace "your_Trace_workspace" \
  --service-name-prefix "your_service_name_prefix"

Windows

# Download the installer script
Invoke-WebRequest `
    -Uri 'https://aliyun-observability-release-cn-shanghai.oss-cn-shanghai.aliyuncs.com/loongsuite-pilot/installer.ps1' `
    -OutFile installer.ps1

# Run the installer. Replace all parameter values with the ones obtained above.
.\installer.ps1 install `
    -CollectLog 'true' `
    -CollectTrace 'true' `
    -SlsProject "your_SLS_Project_name" `
    -SlsLogstore "your_SLS_Logstore_name" `
    -SlsEndpoint "your_SLS_endpoint" `
    -CmsLicenseKey "your_Trace_License_Key" `
    -CmsEndpoint "your_Trace_endpoint" `
    -CmsWorkspace "your_Trace_workspace" `
    -ServiceNamePrefix "your_service_name_prefix"

The installer automatically performs the following operations:

• Detects the AI Coding Agents supported on your machine.

• Allows you to select the agents whose data you want to collect.

• Deploys the corresponding hooks or plugins.

• Writes the local configuration.

• Starts the LoongSuite Pilot background service.

After the installation is complete, run the following commands to check the service status:

loongsuite-pilot status
loongsuite-pilot info

By default, LoongSuite Pilot writes local configuration, runtime status, and JSONL collection results to the .loongsuite-pilot directory under the current user's home directory. The default paths are as follows:

Manage LoongSuite Pilot

The local dashboard starts and stops together with the LoongSuite Pilot collection service. You do not need to run a separate monitor command. After the collection service starts, open the following default address. If you configured a custom port, replace 8765 with that port:

http://127.0.0.1:8765/

To uninstall LoongSuite Pilot and retain local data, run the following command:

macOS / Linux:

curl -fsSL https://aliyun-observability-release-cn-shanghai.oss-cn-shanghai.aliyuncs.com/loongsuite-pilot/installer.sh -o /tmp/loongsuite-pilot-installer.sh && bash /tmp/loongsuite-pilot-installer.sh uninstall

Windows:

.\installer.ps1 uninstall

To remove both installation files and local data, run the following command:

macOS / Linux:

curl -fsSL https://aliyun-observability-release-cn-shanghai.oss-cn-shanghai.aliyuncs.com/loongsuite-pilot/installer.sh -o /tmp/loongsuite-pilot-installer.sh && bash /tmp/loongsuite-pilot-installer.sh uninstall --purge

Windows:

.\installer.ps1 uninstall -Purge

Configure Agent Collection

After installation, you can control whether to collect data from a specific AI Coding Agent by modifying the agents section in the local configuration file. The default configuration file path is as follows:

{
  "agents": {
    "claude-code": {
      "enabled": true
    },
    "cursor": {
      "enabled": true
    },
    "qoder": {
      "enabled": false
    }
  }
}

For example, to collect data from Claude Code and Cursor but disable Qoder, use the configuration shown above. After you modify the configuration, restart LoongSuite Pilot:

loongsuite-pilot restart

Configure Data Masking

LoongSuite Pilot can mask common credentials and personal sensitive data in normalized events before they are sent to the output backend. Masking applies to scenarios in which prompts, completions, tool parameters, or tool results may contain credentials or personal sensitive data.

To enable or adjust the masking policy, modify the local configuration file. The default configuration file path is as follows:

To enable all masking rules:

{
  "mask": {
    "mode": "all"
  }
}

To enable only specific masking rules:

{
  "mask": {
    "mode": "custom",
    "types": [
      "apiKey",
      "idCard",
      "phone",
      "email",
      "ipAddress",
      "bankCard"
    ]
  }
}

The following table describes the masking modes.

The following table describes the masking types.

LoongSuite Pilot primarily scans fields that may contain user or tool content, such as:

• LLM input and output messages.

• Tool call parameters.

• Tool call results.

• Known agent content fields.

Stable metadata such as model names, token counts, latency, Git branches, and workspace paths are not scanned as credential content fields.

After you modify the configuration, run the following command to restart LoongSuite Pilot:

loongsuite-pilot restart

Verify the Onboarding Result

After installation or configuration changes, follow these steps to verify that the onboarding is successful.

loongsuite-pilot status
loongsuite-pilot info

ls ~/.loongsuite-pilot/logs/output
tail -f ~/.loongsuite-pilot/logs/output/*.jsonl

Get-ChildItem "$env:USERPROFILE\.loongsuite-pilot\logs\output"

Get-ChildItem "$env:USERPROFILE\.loongsuite-pilot\logs\output\*.jsonl" |
  Sort-Object LastWriteTime |
  Select-Object -Last 1 |
  Get-Content -Wait

FAQ