DAMASCO Agents
Last updated
Last updated
Note: Damasco Agents is an enterprise feature, allowing you to consolidate, monitor, and configure multiple AI-driven DeFi applications in one place.
Modern DeFi organizations often run multiple AI-powered applications across different workflows and environments. Each app or microservice may serve unique purposes—from user-facing trading bots to internal risk analysis tools—all of which rely on LLM-driven logic and require unique security configurations.
Agents in Damasco make it easy to organize and manage these varied use cases. An Agent can represent a single integration of Damasco or a collection of related integrations—for example, you might have one Agent per application, environment, or major component.
By configuring each Agent separately, you can:
Apply Different Defense Policies Fine-tune your Damasco defenses (Prompt Injection Prevention, Harmful Content Moderation, Data Leakage Controls, and Smart Contract Integrity Checks) for each AI-powered application or environment.
Separate External vs. Internal Tools Keep user-facing DeFi solutions clearly distinct from in-house AI tools, making oversight and troubleshooting simpler.
Compare Testing vs. Production Track performance in development environments independently from production traffic, helping you quickly pinpoint and isolate security or performance issues.
Identify Exploit Activity Determine which part of your system might be under attack by segmenting logs and analytics by Agent.
Click “New Agent”
Provide an Agent Name, which can reflect the application or environment (e.g., “User Trading Bot – Prod,” “Internal QA Tool – Test”).
Optionally add Tags, such as “Application,” “Model,” or any custom key-value labels you want to track (e.g., “Environment:Production”).
Assign a Policy
Select the defense policy this Agent should use. Policies define which detectors to enable, threshold levels, and other security configurations.
If you need a new or customized policy, create it in the Policies page first.
Generate Agent ID
After you save, a unique Agent ID is automatically created for this Agent. All future requests for this Agent must include this ID to ensure the correct policy and metadata are applied.
Tip: You can also manage Agents via the command line or API (self-hosted deployments), where you edit a policy file to add or remove Agents.
Every screening request made to Damasco should include the Agent ID so that logs, analytics, and policy settings map to the correct Agent. If no Agent ID is provided, Damasco uses the Default Policy, which runs all defenses at their default thresholds.
This approach ensures each Agent’s usage and threat data can be filtered or segmented in the Damasco dashboard.
Agents Page: Lists all Agents in your organization along with high-level details, including assigned policies and tags.
Filtering: Use filters (e.g., by tag or Agent Name) to view logs, usage metrics, and threats specific to each Agent.
Updating the Assigned Policy
Click the three dots next to the Agent’s name and select “Edit.”
Choose a new policy from the dropdown. Changes take effect immediately, so proceed with caution to avoid unintended disruptions.
Adding or Modifying Tags
You can add or change metadata tags (e.g., updating “Model:gpt-3.5” to “Model:gpt-4”).
Note: Changing tags is retroactive—past logs will show the updated tag. If you want to preserve historical accuracy, consider creating a new Agent instead.
Ensure No Active Requests Use the Deleted Agent
If your app still references the old Agent ID, requests will fail once that Agent is removed.
Click “Delete Agent”
Confirm the deletion in the popup. This is irreversible—the Agent’s logs remain but are disassociated from any Agent ID or tags.
Past requests keep any request-specific metadata (e.g., user session ID), but lose the Agent’s configuration or tags.
Production vs. Development
Maintain separate Agents for “Prod” and “Dev” environments. Compare flagged prompts or contract anomalies to see if new code in Dev triggers more or fewer warnings before a production release.
Multi-App Organizations
If you have a consumer-facing trading chatbot and an internal compliance auditing tool, assign each its own Agent for distinct policies and analytics.
Compliance & Regulatory Requirements
For strict data protection or advanced DeFi oversight, create specialized Agents with more stringent thresholds (e.g., L4 or “paranoid mode”) to reduce the risk of data leaks or malicious activity.
Frequent Model Upgrades
Spin up a new Agent whenever changing LLM providers or moving from GPT-3.5 to GPT-4, preserving the historical logs for the old model.
Plan Your Tagging Strategy: Use consistent naming for your “Application,” “Environment,” or “Model” tags to compare performance across Agents.
Limit Policy Changes: Rapidly switching an Agent’s policy can cause confusion or inaccurate logs. If you need significantly different settings, consider creating a new Agent to keep historical data intact.
Audit Regularly: Periodically review the Agents dashboard to see if your security posture (flagged interactions, thresholds) meets the evolving needs of each application.
Damasco Agents provide granular control and visibility across all your AI-driven DeFi solutions, enabling you to tailor defense settings to each unique environment or application. By organizing your integrations into Agents—with their own tags, policies, and tracking—you gain a clear picture of security performance, swiftly isolate threats, and maintain precise oversight in a complex, multi-application landscape.