This privacy agreement explains how AgentLoop collects, evaluates, audits, and stores your data. AgentLoop does not use your data for any purpose other than those described in this agreement.
Data collection
AgentLoop collects data through the following components.
Probe
The collection code is open source.
Across the entire pipeline, data can be reported over TLS. Unless you specify otherwise, your data is not written to local disk or processed locally in any other way. It is reported directly to Simple Log Service for storage.
LoongSuite-Pilot
LoongSuite-Pilot supports AI agents such as Claude Code, Codex, Cursor, and Qoder. For the supported scope, see the integration center. You can enable data collection for specific AI agents as needed, or enable automatic detection so that collection starts as soon as a matching data source is detected.
The core collection code of LoongSuite-Pilot is open source, so you can review the collection scope and the processing logic.
If your AI agent processes sensitive information, such as personal information, code, or credentials, during model calls or tool calling, LoongSuite-Pilot collects the related data based on the calls that actually occur. This data is used for AI agent observability, usage analysis, troubleshooting, and security auditing. An administrator can configure masking rules to process the data.
LoongCollector
LoongCollector runs based on the collection pipeline configured by the administrator, who can specify collection sources, processing rules, and output destinations.
After you enable the AgentSight collection plug-in, if your AI agent processes sensitive information, such as personal information, code, or credentials, during model calls or tool calling, LoongCollector collects the related data based on the calls that actually occur. An administrator can process the data by using masking rules.
Evaluation
The model provider used by the evaluation agent is Alibaba Cloud Model Studio. According to Security certifications and privacy, Model Studio commits that data is not used for model training.
Auditing
After you enable model semantic detection, AgentLoop submits the session context related to a risk event to Alibaba Cloud Model Studio for risk assessment. According to Security certifications and privacy, Model Studio commits that the related data is not used for model training.
Audit data is visible only to users who have the required permissions.
Storage
Storage location — The data of AgentLoop users is stored in Simple Log Service under the user's own account. The scope of data usage is controlled through RAM permissions.
Encryption — Simple Log Service supports KMS-encrypted storage. You can enable KMS encryption to protect the privacy of your data.
Privacy statement — For the Simple Log Service privacy statement, see Security and compliance.