KYA™ · Know Your Agent

Agent Trust Registry

Public registry of preliminary KYA™ signals for source-visible agent projects.
Scores support due diligence and capability controls. They are not final security certifications.

30
Agents tracked
30
Preliminary scans
2026
Findings flagged
765
CVEs detected
Static analysis + OSV.dev CVE scan + NVIDIA Llama 70B audit · Questions? [email protected]
Tiers
Sovereign ≥85
Partner ≥72
Node ≥58
Rejected <58
Blended score: 30% manual baseline + 70% live analysis
Audit Methodology

How we score agents

Every score in this registry is produced by a three-stage preliminary pipeline run against a prioritized sample of the agent's public source code — no installs, no clones, no marketing materials. We inspect capabilities, check dependencies, and run an independent AI review.

Scores are a blend of our manual safety baseline (30%) and the live analysis result (70%). The baseline captures things code can't show: organizational maturity, incident history, and published safety disclosures. The live analysis reflects what's actually in the codebase today.

On false positives. Static analysis is inherently noisy. A CLI tool using child_process looks the same as a malicious subprocess call. An agent that legitimately browses the web will flag network patterns. We surface these signals — we don't suppress them — because the operator needs to make that judgment for their context. Over time, our detection patterns improve as we build type-aware and context-aware rules. Treat scores as a starting point for due diligence, not a final verdict.

01

Static code analysis

Pattern-based scan across a prioritized source sample fetched via GitHub API — no disk writes, no execution. We flag capabilities such as shell execution, code evaluation, and network access for review. A capability signal is not, by itself, a confirmed vulnerability.

02

Dependency CVE scan

We parse package.json and requirements.txt files without installing anything. Each dependency is queried against OSV.dev — Google's open vulnerability database — using their batch API. CVSS ≥7.0 is flagged as High, 4.0–7.0 as Medium. We also flag unpinned version ranges (^, ~, *) as supply chain risk.

03

AI-powered semantic audit

Source files are passed to a large language model (Llama 3.1 70B via NVIDIA NIM) with a structured security prompt. The model looks for issues static patterns miss: unsafe prompt construction, missing input validation on tool calls, context leakage between sessions, and missing human-in-the-loop checkpoints. This stage catches behavioral risks, not just syntactic ones.

04

Score composition

Five dimensions are scored: Framework (design-level guardrails), Code Health (quality and safety of implementation), Tool Permissions (blast radius of tool access), Prompt Safety (injection resistance), and Loop Safety (termination guarantees). These combine into a raw score, which is then blended with our manual baseline. Scores are re-run periodically as frameworks evolve.

Code Source Available
79
Claude Code
Anthropic
Partner Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
H:2 M:2 I:79
Prompt SafetyTool Abuse Risk
Top finding ████████████████████████████████████ — classified
20/26 prioritized files · 1c5f951
General Open Source
78
IronClaw
Near AI
Partner Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
H:1 M:2 I:168
Prompt SafetyTool Abuse Risk
Top finding ████████████████████████████████████ — classified
13 deps scanned via OSV.dev
24/384 prioritized files · 0377180
General Open Source
77
OpenClaw
OpenClaw Labs
Partner Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
H:1 M:2 I:2
Prompt SafetyTool Abuse Risk
Top finding ████████████████████████████████████ — classified
56 deps scanned via OSV.dev
60/16364 prioritized files · aeb537c
General Open Source
76
LangChain Agent
LangChain Inc.
Partner Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
H:1 M:2 I:1
Prompt Safety
Top finding ████████████████████████████████████ — classified
60/2490 prioritized files · 23ce677
General Open Source
76
PicoClaw
Sipeed
Partner Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
H:1 M:2 I:5
Prompt SafetyTool Abuse Risk
Top finding ████████████████████████████████████ — classified
50 deps scanned via OSV.dev
60/64 prioritized files · b9a8fad
Multi-Agent Open Source
74
LangGraph
LangChain Inc.
Partner Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
H:2 M:1 I:6
Prompt SafetyTool Abuse Risk
Top finding ████████████████████████████████████ — classified
20 deps scanned via OSV.dev
60/461 prioritized files · d57a74f
General Open Source
74
UI-TARS
ByteDance
Partner Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
H:2 M:2 I:5
Prompt SafetyTool Abuse Risk
Top finding ████████████████████████████████████ — classified
4/5 prioritized files · 582f3a7
General Open Source
73
ZeroClaw
ZeroClaw Labs
Partner Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
H:2 M:2 I:37
Prompt SafetyTool Abuse Risk
Top finding ████████████████████████████████████ — classified
26 deps scanned via OSV.dev
44/47 prioritized files · bab4da8
Multi-Agent Open Source
69
CrewAI
CrewAI Inc.
Node Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
H:2 M:2 I:41
Prompt SafetyTool Abuse Risk
Top finding ████████████████████████████████████ — classified
60/1228 prioritized files · f214ff4
Multi-Agent Open Source
69
AutoGen
Microsoft Research
Node Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
H:2 M:3 I:9 CVE:1
Dependency VulnerabilityPrompt SafetyTool Abuse Risk
Top finding ████████████████████████████████████ — classified
38 deps scanned via OSV.dev
60/593 prioritized files · 027ecf0
General Open Source
68
PydanticAI
Pydantic
Node Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
H:3 M:2 I:5
Loop SafetyPrompt SafetyTool Abuse Risk
Top finding ████████████████████████████████████ — classified
6 deps scanned via OSV.dev
49/613 prioritized files · 13931c1
Research Open Source
65
Dexter
virattt
Node Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
H:2 M:5 I:12 CVE:3
Dependency VulnerabilityPrompt SafetyTool Abuse Risk
Top finding ████████████████████████████████████ — classified
35 deps scanned via OSV.dev
60/185 prioritized files · 4adf938
General Open Source
63
Moltis
Moltis Org
Node Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
M:12 I:27 CVE:10
Dependency VulnerabilityPrompt SafetyTool Abuse Risk
23 deps scanned via OSV.dev
60/272 prioritized files · 48c9a41
Multi-Agent Open Source
62
OpenAI Swarm
OpenAI
Node Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
H:4 M:3 I:4
Loop SafetyPrompt SafetyTool Abuse Risk
Top finding ████████████████████████████████████ — classified
55/63 prioritized files · 6af0b4c
General Open Source
58
LlamaIndex
LlamaIndex Inc.
Node Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
H:1 M:12 I:1 CVE:10
Dependency VulnerabilityTool Abuse Risk
Top finding ████████████████████████████████████ — classified
17 deps scanned via OSV.dev
60/3958 prioritized files · 363d0b7
General Open Source
56
Semantic Kernel
Microsoft
Rejected Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
H:2 M:82 I:11 CVE:81
Loop SafetyDependency VulnerabilityPrompt Safety
Top finding ████████████████████████████████████ — classified
34 deps scanned via OSV.dev
60/1266 prioritized files · 61331d8
General Open Source
56
UI-TARS Desktop
ByteDance
Rejected Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
H:2 M:154 I:63 CVE:152
Dependency VulnerabilityPrompt SafetyTool Abuse Risk
Top finding ████████████████████████████████████ — classified
376 deps scanned via OSV.dev
60/1252 prioritized files · e9f3387
General Open Source
55
Mastra
Mastra AI
Rejected Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
H:2 M:79 I:54 CVE:77
Dependency VulnerabilityPrompt SafetyTool Abuse Risk
Top finding ████████████████████████████████████ — classified
187 deps scanned via OSV.dev
60/6079 prioritized files · cff40d4
Multi-Agent Open Source
55
Paperclip
PaperclipAI
Rejected Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
H:2 M:19 I:37 CVE:17
Dependency VulnerabilityPrompt SafetyTool Abuse Risk
Top finding ████████████████████████████████████ — classified
99 deps scanned via OSV.dev
60/1579 prioritized files · f3db7b8
General Open Source
55
OpenFang
RightNow AI
Rejected Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
H:2 M:12 I:392 CVE:10
Dependency VulnerabilityPrompt SafetyTool Abuse Risk
Top finding ████████████████████████████████████ — classified
9 deps scanned via OSV.dev
36/39 prioritized files · acf2587
General Open Source
54
Agno
Agno (ex-Phidata)
Rejected Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
H:2 M:131 I:3 CVE:129
Dependency VulnerabilityPrompt SafetyTool Abuse Risk
Top finding ████████████████████████████████████ — classified
354 deps scanned via OSV.dev
60/3948 prioritized files · 3adb736
General Open Source
53
AutoGPT
Significant Gravitas
Rejected Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
H:1 M:14 I:21 CVE:11
Dependency VulnerabilityPrompt SafetyTool Abuse Risk
Top finding ████████████████████████████████████ — classified
150 deps scanned via OSV.dev
20/1865 prioritized files · 3417f13
General Open Source
52
Hermes Agent
NousResearch
Rejected Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
H:2 M:19 I:18 CVE:17
Dependency VulnerabilityPrompt SafetyTool Abuse Risk
Top finding ████████████████████████████████████ — classified
159 deps scanned via OSV.dev
36/2798 prioritized files · 07ac185
Code Open Source
50
OpenHands
All Hands AI
Rejected Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
H:5 M:7 I:10 CVE:5
Loop SafetyDependency VulnerabilityPrompt SafetyTool Abuse Risk
Top finding ████████████████████████████████████ — classified
96 deps scanned via OSV.dev
60/1321 prioritized files · 94c60d7
General Open Source
50
smolagents
Hugging Face
Rejected Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
H:2 M:84 I:35 CVE:81
Loop SafetyDependency VulnerabilityPrompt SafetyTool Abuse Risk
Top finding ████████████████████████████████████ — classified
26 deps scanned via OSV.dev
27/77 prioritized files · e8b988d
General Open Source
50
Cherry Studio
CherryHQ
Rejected Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
H:3 M:63 I:10 CVE:61
Loop SafetyDependency VulnerabilityPrompt SafetyTool Abuse Risk
Top finding ████████████████████████████████████ — classified
418 deps scanned via OSV.dev
60/2529 prioritized files · 8d8de95
General Open Source
48
Nanobot
HKUDS
Rejected Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
H:2 M:19 I:30 CVE:17
Loop SafetyDependency VulnerabilityTool Abuse RiskPrompt Safety
Top finding ████████████████████████████████████ — classified
52 deps scanned via OSV.dev
34/462 prioritized files · b8a4ceb
Multi-Agent Open Source
47
MetaGPT
FoundationAgents
Rejected Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
H:3 M:57 I:13 CVE:55
Loop SafetyDependency VulnerabilityPrompt SafetyTool Abuse Risk
Top finding ████████████████████████████████████ — classified
97 deps scanned via OSV.dev
60/919 prioritized files · 11cdf46
Multi-Agent Open Source
47
Agency Swarm
VRSEN
Rejected Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
H:2 M:23 I:13 CVE:21
Loop SafetyDependency VulnerabilityTool Abuse RiskPrompt Safety
Top finding ████████████████████████████████████ — classified
22 deps scanned via OSV.dev
60/316 prioritized files · dad735a
General Open Source
43
Strands Agents
AWS / Strands
Rejected Preliminary scan Jun 10, 2026 Static refresh · AI findings preserved
H:4 M:10 I:25 CVE:7
Loop SafetyDependency VulnerabilityPrompt SafetyTool Abuse Risk
Top finding ████████████████████████████████████ — classified
119 deps scanned via OSV.dev
60/1060 prioritized files · 3db1b63

Apply for a KYA™ evaluation.

Approved capabilities depend on your agent, operator, jurisdiction, cohort, and licensed infrastructure partners. A preliminary registry score does not guarantee financial access.

Apply for Access