Supported Frameworks
Inkog detects logic flaws and security risks across 16+ AI agent frameworks with consistent results, powered by our Universal IR architecture.
Supported Frameworks
Feature Support Matrix
All frameworks receive the same analysis through Inkog’s Universal IR. Here’s the feature breakdown:
| Framework | Scan | AGENTS.md | Compliance | Topology |
|---|---|---|---|---|
| LangChain | ✅ | ✅ | ✅ | ✅ |
| LangGraph | ✅ | ✅ | ✅ | ✅ |
| CrewAI | ✅ | ✅ | ✅ | ✅ |
| AutoGen (AG2) | ✅ | ✅ | ✅ | ✅ |
| DSPy | ✅ | ✅ | ✅ | ✅ |
| Phidata | ✅ | ✅ | ✅ | ✅ |
| Smolagents | ✅ | ✅ | ✅ | ✅ |
| LlamaIndex | ✅ | ✅ | ✅ | ✅ |
| n8n | ✅ | ✅ | ✅ | ✅ |
| Flowise | ✅ | ✅ | ✅ | ✅ |
| Langflow | ✅ | ✅ | ✅ | ✅ |
| Dify | ✅ | ✅ | ✅ | ✅ |
| Google ADK | ✅ | ✅ | ✅ | ✅ |
| Copilot Studio | ✅ | ✅ | ✅ | ✅ |
| Agentforce | ✅ | ✅ | ✅ | ✅ |
| OpenAI Agents | ✅ | ✅ | ✅ | ✅ |
| Azure AI Foundry | ✅ | ✅ | ✅ | ✅ |
Legend:
- Scan - Full static analysis (25+ patterns)
- AGENTS.md - Governance manifest validation
- Compliance - EU AI Act, NIST, OWASP, ISO 42001 mapping
- Topology - Agent workflow graph generation
Code-First Frameworks
Build agents programmatically with full control over behavior and logic.
| Framework | Language | Detection | Docs |
|---|---|---|---|
| LangChain | Python | 25 patterns | View → |
| LangGraph | Python | 25 patterns | View → |
| CrewAI | Python | 25 patterns | View → |
| AutoGen (AG2) | Python | 25 patterns | View → |
| DSPy | Python | 25 patterns | View → |
| Phidata | Python | 25 patterns | View → |
| Smolagents | Python | 25 patterns | View → |
No-Code / Low-Code Platforms
Visual builders and workflow automation tools with JSON/YAML configurations.
| Platform | Config Type | Detection | Docs |
|---|---|---|---|
| n8n | JSON Workflows | 25 patterns | View → |
| Flowise | JSON Workflows | 25 patterns | View → |
| Langflow | JSON Workflows | 25 patterns | View → |
| Dify | YAML/JSON | 25 patterns | View → |
Enterprise Platforms
Official SDKs and enterprise-grade agent frameworks.
| Platform | Provider | Detection | Docs |
|---|---|---|---|
| Google ADK | 25 patterns | View → | |
| Copilot Studio | Microsoft | 25 patterns | View → |
| Agentforce | Salesforce | 25 patterns | Coming Soon |
| OpenAI Agents | OpenAI | 25 patterns | View → |
| Semantic Kernel | Microsoft | 25 patterns | View → |
| Azure AI Foundry | Microsoft | 25 patterns | View → |
RAG Frameworks
Retrieval-Augmented Generation frameworks for knowledge-enhanced agents.
| Framework | Language | Detection | Docs |
|---|---|---|---|
| LlamaIndex | Python | 25 patterns | View → |
| Haystack | Python | 25 patterns | View → |
Universal IR Architecture
Unlike tools that require separate parsers per framework, Inkog uses a Universal Intermediate Representation (Universal IR):
Your Code → Framework Adapter → Universal IR → Pattern Detection → FindingsBenefits:
- Write once, detect everywhere - Same detection rules work across all frameworks
- Consistent results - Identical vulnerabilities produce identical findings
- Future-proof - Adding new frameworks only requires an adapter, not rewriting detection logic
Learn more about the Universal IR architecture →
Detection Coverage
All frameworks receive the same comprehensive security analysis:
| Category | Patterns | Examples |
|---|---|---|
| Prompt Security | 5 | Prompt injection, SQL injection via LLM |
| Code Execution | 4 | Unsafe eval, pickle, YAML loading |
| Resource Exhaustion | 5 | Infinite loops, token bombing, context overflow |
| Credential Exposure | 3 | Hardcoded secrets, unsafe env access |
| Governance | 4 | Missing oversight, authorization, audit logging |
| Data Security | 4 | Cross-tenant leakage, logging sensitive data |
Request Framework Support
Don’t see your framework? Open an issue on GitHub or contact us at support@inkog.io.