v1.7.0 — 147 tools now indexed|PromptArmor, HiddenLayer, Patronus AI + 10 more →
Open Source · March 2026

The Security Index for Generative & Agentic AI

147 tools mapped to the OWASP LLM Top 10 and Agentic Top 10. Search by risk, ask in plain English, or explore the knowledge graph.

147
Curated Tools
11
Categories
20
OWASP Risks
214
Tags
⌘K Ask in plain English... Enter
Example: "tools for prompt injection in agentic workflows"
LLM Guard (Protect AI)
AI Guardrails · LLM01, ASI04
Plug & Play
Gandalf (Lakera)
AI Red Teaming · LLM01, LLM02
Training
Promptfoo
AI Red Teaming · LLM01, LLM07
Open Source
The Problem

Why are OWASP checklists not enough for production AI?

The frameworks exist. The industry knows the risks. But most teams are stuck between a PDF and a spreadsheet — with no path from "address LLM06" to "here's the tool."

Without Yuntona
  • OWASP Top 10 lives in a PDF nobody reopens
  • Tool discovery via word-of-mouth and conferences
  • No structured mapping from risk to remediation
  • Coverage gaps invisible until audit or incident
With Yuntona
  • Every risk links to the tools that address it
  • AI search returns ranked results in seconds
  • Knowledge graph allows visual exploration of risk coverage
  • Continuously updated as new tools emerge
Platform

How does Yuntona help you find AI security tools?

Three views designed for different workflows — browse, search, or explore connections.

How it works

How do you find the right tool?

Start from a risk, a category, or a plain-English question.

Natural language search

Type a question and Yuntona's search parses your intent, extracts filters, and returns ranked results from 147 tools. A future iteration will add conversational follow-up queries.

→ "open source tools for securing MCP integrations"
category: MCP Security tag: open source risk: ASI04
3 results · Stacklok/ToolHive, MCP Secure Gateway, FuzzingLabs

See the whole landscape

The knowledge graph visualises every tool, risk, and lifecycle stage as an interactive network. Click any node to highlight its connections — revealing coverage clusters and defence gaps at a glance.

OWASP Mapping

What OWASP risks does Yuntona cover?

All 20 risk categories — with real tool counts per risk.

RiskNameToolsCoverage
LLM01Prompt Injection53
LLM02Sensitive Information Disclosure46
LLM03Supply Chain Vulnerabilities19
LLM04Data and Model Poisoning13
LLM05Improper Output Handling18
LLM06Excessive Agency51
LLM07System Prompt Leakage41
LLM08Vector & Embedding Weaknesses37
LLM09Misinformation19
LLM10Unbounded Consumption8
RiskNameToolsCoverage
ASI01Excessive Agency & Autonomy53
ASI02Insufficient Access Controls51
ASI03Inadequate Sandboxing32
ASI04Insecure Tool Integration41
ASI05Insecure Output Consumption25
ASI06Lack of Runtime Guardrails20
ASI07Insecure Multi-Agent Orchestration26
ASI08Inadequate Logging & Monitoring16
ASI09Trust Boundary Violations16
ASI10Insufficient Error Handling17
Curator

Who built Yuntona?

Expert curation means accountability.

FB
Fabio Baumeler
Third-Party Cyber Risk Lead · Financial Conduct Authority

Security practitioner with over a decade of experience spanning SOC/NOC operations, business information security, and financial services regulation. Built Yuntona to close the gap between AI security frameworks and the tools that operationalise them.

CISSPMSc InfoSec · Royal HollowayGCHQ-certified
Common Questions

What do security teams ask about Yuntona?

Quick answers to the questions practitioners and AI search engines ask most.

How do you decide what tools go into the directory?

Every tool is individually evaluated against three criteria: it addresses a genuine security risk in the generative or agentic AI stack, it is operational or near-operational (not vaporware), and it offers meaningful capability not already covered by existing entries. Discovery draws from OWASP working groups, industry conferences, practitioner networks, and primary research. Tools that duplicate existing coverage without differentiation are excluded. Read the full methodology →

Is Yuntona free and open source?

Yes. Fully open source under the MIT licence. 147 tools accessible without login or paywall. Source code on GitHub.

How is Yuntona different from an awesome-list or analyst report?

Awesome-lists are flat link collections with no risk mapping. Analyst reports are paywalled snapshots. Yuntona sits between: every tool is individually evaluated, mapped to OWASP risk codes, tagged by complexity, and searchable via AI-powered natural language. Continuously updated.

How does Yuntona handle AI-assisted curation?

Tool discovery is practitioner-led. OWASP risk mappings use AI as an analytical engine — the methodology and schema are human-designed, every output validated against published standards. Human-directed analysis at scale, not automated classification.

Who maintains Yuntona and why should I trust it?

Yuntona is built and maintained by Fabio Baumeler, a CISSP-certified Third-Party Cyber Risk Lead at the UK Financial Conduct Authority with an MSc in Information Security from Royal Holloway (GCHQ-certified). Single-maintainer accountability means every assessment reflects one expert's informed judgement — not crowd-sourced voting or vendor self-submission.

Community

How can you contribute to Yuntona?

Built in the open. Suggest tools, report issues, or contribute directly.

Find the capabilities you're missing

Close the blind spots in your AI risk coverage. 147 tools. 20 OWASP risks. One search.

Explore Directory →
OWASP LLM Top 10 (2025)
OWASP Agentic Top 10 (2026)
147 Tools Curated
Open Source · MIT Licence