Overview

The AI Agent in DeepTraQ is an AI-powered cybersecurity assistant that combines real-time web search, vulnerability analysis, automated patch generation, database interrogation, and full scan orchestration into a single workflow. From a single UI you can select an AI engine, ask the agent about a CVE (for example, CVE-2024-74), validate the agent’s sources via the References tab, generate patch scripts, query your environment database for presence of the vulnerability, and create or invoke DeepTraQ scans — all without leaving the platform.

Supported Platforms

• Code and infrastructure vulnerability workflows that surface CVE-style findings.
• Model-agnostic AI engines: ChatGPT, Claude, Gemini (and other configured engines via the model selector).
• Real-time web search integration for grounding AI responses (security blogs, exploit databases, CVE pages).
• Internal database integrations (the "DB" button to query your environment inventory).
• DeepTraQ scan orchestration and web-application security dashboards (scan creation, invocation, and status monitoring).

Prerequisites

Requirement	Description	Example / Notes
Configured AI engines	At least one AI model must be registered and available in DeepTraQ's model selector.	ChatGPT, Claude, Gemini configured in model settings.
Web search tool enabled	Allow the agent to perform real-time internet lookups to ground responses and cite sources.	Required to fetch exploit write-ups or CVE references in the References tab.
Database access configured	Credentials and read access to the inventory/asset database so the DB button can query presence of CVEs.	Connection to your asset DB with a service account that can run vulnerability queries.
Scan integration enabled	DeepTraQ AI scans must be integrated so the agent can create and invoke scans from the agent UI.	Integration with open-source scanners + DeepTraQ proprietary scan engine.
User permissions	User must have permission to view findings, run scans, and (optionally) apply generated patch scripts.	Role with Scan:invoke, DB:query, Findings:view privileges.

Steps

1. Open the AI Agent module inside DeepTraQ.
2. Use the model selector to choose an AI engine (for example, ChatGPT, Claude, or Gemini).
3. Ask the agent a targeted question — e.g., “Tell me about CVE-2024-74: root cause and impact.”
4. Review the agent’s generated response in the UI.
5. Open the References tab to verify the sources the agent used and ensure there are direct links to external resources (security blogs, exploit DB, CVE pages).
6. (Optional) Ask the agent to generate a patch script for the CVE.
7. Click the DB button to query your internal database for the presence of the CVE in your environment.
8. If a vulnerability exists, instruct the agent to create and invoke an AI scan or a targeted scan from the agent UI.
9. Monitor scan progress and results via the Web Application Security Dashboard or the scan status panel.
10. Validate findings by following source links from the References tab and, if needed, iterate with follow-up agent prompts (root cause, remediation steps, or patch adjustments).

Field Reference

Field	Description	Example
Model selector	UI control to pick which AI engine the agent will use for analysis.	Select "Gemini" or "ChatGPT" before asking about a CVE.
Agent response panel	Primary display area where the agent’s structured analysis is shown.	"Root cause: vulnerable library X; Impact: RCE risk."
References tab	Lists external sources the agent used and provides direct links to each source to prevent hallucinations.	Link to a CVE entry, exploit write-up, or vendor advisory.
DB button	Triggers a query against your configured database to check whether the CVE exists in your environment.	"No matches found" or a list of affected hosts.
Patch script generator	Agent command/output that produces a candidate patch or remediation script for the CVE.	Bash script to upgrade library X to version Y.
Scan creation / orchestration	Agent-driven action to create and invoke DeepTraQ-integrated scans combining OSS and proprietary detectors.	Create a web-app scan targeting host list returned by the DB query.
Web Application Security Dashboard	Dashboard where scan progress and results are shown after the agent creates or triggers a scan.	Scan status: queued / running / completed; findings list.
Source links	Clickable links in References that point to the exact web pages used to generate the agent’s intelligence.	External security blog or exploit-db link cited by the agent.
User permissions indicator	Visual hint or requirement that shows necessary permissions for actions (scan invoke, DB query, apply patch).	Permission badges shown next to action buttons.