VAAST
Vulnerability Assessment for AI Security Testing
VAAST is the offensive-security platform for LLM-integrated applications. Map the surface, run the catalog, ship the findings.
Capabilities
Everything the attack surface demands.
Prompt Injection Catalog
Direct, indirect, and multi-turn variants with reproducible payloads. Covers system-prompt exfiltration, jailbreaks, and role-confusion attacks.
- Direct and indirect injection vectors
- Multi-turn conversation exploitation
- System prompt exfiltration techniques
Tool-Call Surface Mapping
VAAST's MCP Scanner enumerates callable tools and probes each one for parameter injection, privilege escalation, schema confusion, and unauthorized state changes.
- Automatic tool enumeration via MCP
- Parameter injection and type confusion
- Authorization boundary testing
Proxy Intelligence
Intercept LLM API traffic and send it to Xtriforce for instant security analysis. API keys in URLs, system prompt leaks, unsafe model parameters — caught automatically.
- HTTP/S interception of LLM API traffic
- Real-time credential and leak detection
- Request/response security analysis
Exportable Evidence
Every finding includes the exact payload and the model's response, exportable as JSON or HTML.
- JSON export on every plan
- Full HTML report on Pro
- Evidence includes exact payload and response
RAG Corpus Inspection
Walks ingestion paths, tests for poisoned documents, and verifies retrieval behavior under adversarial queries.
Agentic Pipeline Tracing
Models plan-and-execute loops and long-horizon agents, then surfaces failure modes unique to autonomous workflows.
Authenticated Scanning
Run the catalog behind API key, Bearer, or Cookie authentication so you can test what your users actually see.
Workflow
How an engagement runs.
Define scope
Register the target application with written authorization. VAAST will refuse to run without it.
Enumerate surface
Map tool calls, retrieval sources, and agent pipelines. VAAST produces a surface graph you can export.
Probe and report
Run the vulnerability catalog. Every finding is triaged, tagged, and exportable as JSON or HTML.
Demo
See it in action.
Real scans. Real findings. No setup required.
Xtriforce
Your AI security analyst, built into VAAST.
Xtriforce reads your scan findings, explains what landed and why, suggests follow-on tests, and writes report narratives — all without leaving the app. Available on Pro and Enterprise.
Finding Analysis
Xtriforce explains each finding in plain language: what the payload triggered, why it's a risk, and how to remediate it.
Report Narratives
Xtriforce writes the findings narrative for your scan report, ready to export as HTML or JSON.
Proxy Intelligence
See above — intercept LLM API traffic and analyze it for security issues in real time.
Bring VAAST into MCP-compatible AI agents and clients. Read-only access to your local scan findings and workspaces.
npx -y @xtrinel/vaast-mcpPricing
Start free. Upgrade when you need depth.
Every tier includes the baseline payload library. Pro and Enterprise add live research sync, full report export, and authenticated scanning.
Kick the tires on VAAST.
- VAAST desktop app (Mac, Windows, Linux)
- Baseline payload library (15 checks)
- Prompt injection, tool-call, RAG, agentic categories
- MCP Scanner (tool enumeration + adversarial probing)
- Local scan history and workspaces
- JSON findings export
- Community research feed access
For individual practitioners.
- Everything in Free
- Xtriforce AI Analyst (findings analysis, attack scenarios, remediation)
- Live payload library sync (new research checks as published)
- Full report export (HTML + JSON)
- Authenticated scanning (API key, Bearer, Cookie)
- Direct email support
For teams running VAAST together.
- Everything in Pro
- Custom seating for your team
- SSO via SAML 2.0 supported
- Org-level usage dashboard with audit logs
- Member activity tracking — who scanned what and when
- Onboarding call included
New to AI red-teaming? Learn it hands-on in the Infirmary — free labs, no login required.
Start learning →