inkog

<p align="center"> <img src="logo.png" width="200" alt="Inkog"> </p> <h3 align="center">The pre-flight check for AI agents.</h3> <p align="center"> Static analysis that catches the bugs only agent code can have — token bombing, prompt injection, missing oversight, compliance gaps — before they ship. </p> <p align="center"> <a href="https://github.com/inkog-io/inkog/releases"><img src="https://img.shields.io/github/v/release/inkog-io/inkog?label=release" alt="Release"></a> <a href="LICENSE"><img src="https://img.shields.io/badge/license-Apache%202.0-blue.svg" alt="License"></a> <a href="https://goreportcard.com/report/github.com/inkog-io/inkog"><img src="https://goreportcard.com/badge/github.com/inkog-io/inkog" alt="Go Report Card"></a> <a href="https://github.com/inkog-io/inkog/actions/workflows/ci.yml"><img src="https://github.com/inkog-io/inkog/actions/workflows/ci.yml/badge.svg" alt="CI"></a> <a href="https://discord.gg/NuG4SSGRH"><img src="https://img.shields.io/badge/Discord-Join-5865F2?logo=discord&logoColor=white" alt="Discord"></a> </p> <p align="center"> <img src="demo.gif" width="800" alt="Inkog scanning AI agent code for vulnerabilities"> </p> <p align="center"> <a href="https://app.inkog.io/report/83556081-478e-4c64-af52-126b8c83eb6d"> <img src="docs/screenshots/dashboard-report.png" width="800" alt="Inkog dashboard report — findings, code-level analysis, EU AI Act / NIST / OWASP compliance mapping"> </a> <br> _{<i>Every scan is a shareable report — <a href="https://app.inkog.io/report/83556081-478e-4c64-af52-126b8c83eb6d">view this one live</a></i>} </p> --- ## Why Inkog Most security tools find SQL injection. Inkog finds the things that **only break in agent code**: - **Token bombing** — loops where the LLM controls termination, draining your API budget - **Recursive tool calling** — one user request fans out into 10,000 LLM invocations - **Prompt injection sinks** — RAG output flowing into a `system` prompt no one reviewed - **Missing oversight** — destructive tools (refunds, deletes, money) firing without human approval - **Cross-tenant leakage** — global state shared between agent invocations - **MCP tool poisoning** — malicious tool descriptions hijacking your agent Findings map directly to **EU AI Act Article 14 / 15**, **NIST AI RMF**, **ISO 42001**, and **OWASP LLM Top 10** — at the article level, not just bucket labels. [**Try it on a public repo →**](https://inkog.io/scan) · no signup, results in 60 seconds. ## State of AI Agent Security 2026 **We scanned 500+ open-source AI agents. 85% had at least one vulnerability.** Other findings from the data: - **63%** of findings were CRITICAL or HIGH severity - **25%** failed EU AI Act Article 14 (human oversight) - **11,705** total findings across LangChain · CrewAI · AutoGen · pydantic-ai · LangGraph · MCP servers · OpenAI Agents · n8n · Flowise · DSPy The largest security analysis of the AI agent ecosystem — original data from automated static analysis, not surveys. [**Read the full report →**](https://inkog.io/report) ## Quick start ```bash # No install npx -y @inkog-io/cli scan . # Or install permanently brew tap inkog-io/inkog && brew install inkog go install github.com/inkog-io/inkog/cmd/inkog@latest ``` ```bash # Get a free API key at app.inkog.io export INKOG_API_KEY=sk_live_... inkog . ``` ## GitHub Actions ```yaml - uses: inkog-io/inkog@v1 with: api-key: ${{ secrets.INKOG_API_KEY }} sarif-upload: true # findings show in the GitHub Security tab ``` [Workflow example](examples/ci/github-actions.yml) · [GitLab / Azure / Jenkins templates](docs/CI_CD_INTEGRATION.md) ## How Inkog compares The closest direct alternative to Inkog is **[SplxAI Agentic Radar](https://github.com/splx-ai/agentic-radar)** — also OSS, also static analysis of agent code. Honest side-by-side: | | **Inkog** | SplxAI Agentic Radar | |---|---|---| | Approach | Static code analysis | Static code analysis | | Frameworks supported | 21 (Python · TS · no-code) | 4 (CrewAI · LangGraph · OpenAI Agents · n8n) | | Compliance mapping | Article-level (EU AI Act, NIST, ISO 42001, OWASP) | Generic risk taxonomy | | MCP server auditing | ✓ | – | | AGENTS.md governance verification | ✓ | – | | Topology visualization | – | ✓ (interactive graph) | | GitHub stars | 28 | 956 | | License | Apache 2.0 CLI · proprietary engine | Fully OSS | **Different problem, complementary tools** — use Inkog *with* one of these, not instead of: - **Dev-environment scanning** — [Snyk Agent Scan](https://github.com/snyk/agent-scan), [AgentShield](https://github.com/affaan-m/agentshield) audit installed MCP servers and editor configs on your laptop (different scan target — your laptop, not your repo) - **Runtime adversarial probing** — [Lakera](https://www.lakera.ai/ai-red-teaming), [Straiker](https://www.straiker.ai/), [Crucible](https://crucible-security.github.io/crucible-website/), [MS Red Teaming Agent](https://learn.microsoft.com/en-us/azure/foundry/concepts/ai-red-teaming-agent), [NVIDIA Garak](https://github.com/NVIDIA/garak) test deployed agents at the API boundary - **Quality / hallucination evaluation** — [Giskard](https://github.com/Giskard-AI/giskard-oss), [Patronus AI](https://www.patronus.ai/) test answer correctness and safety, not code-level security [Detailed comparison →](docs/comparisons.md) ## Frameworks **Code-first:** LangChain · LangGraph · CrewAI · AutoGen · AG2 · OpenAI Agents · Semantic Kernel · Azure AI Foundry · LlamaIndex · Haystack · DSPy · Phidata · Smolagents · PydanticAI · Google ADK **No-code:** n8n · Flowise · Langflow · Dify · Microsoft Copilot Studio · Salesforce Agentforce ## Use from your editor ```bash npx -y @inkog-io/mcp ``` Adds Inkog as an MCP server in Claude Code, Cursor, ChatGPT — 7 tools including MCP server auditing, Skill package scanning, multi-agent topology analysis. [MCP integration →](https://docs.inkog.io/integrations/mcp) ## More features <details><summary><strong>Deep scan</strong> — orchestrator-driven analysis with enriched findings, agent profile, HTML report</summary> ```bash inkog -deep . inkog -deep -output html . > report.html ``` [Deep scan docs →](https://docs.inkog.io/cli/deep-scan) </details> <details><summary><strong>Skill & MCP scan</strong> — audit SKILL.md packages and MCP servers</summary> ```bash inkog skill-scan . inkog mcp-scan github inkog skill-scan --deep --repo https://github.com/org/repo ``` [Skill & MCP scan docs →](https://docs.inkog.io/cli/skill-scan) </details> <details><summary><strong>Inkog Red</strong> — adversarial testing of running agents</summary> ```bash inkog red --target https://your-agent.example.com ``` Probes prompt injection, jailbreaks, and tool misuse against live endpoints. [Inkog Red docs →](https://docs.inkog.io/red) </details> <details><summary><strong>Scan policies</strong> — five presets from low-noise to full-audit</summary> ```bash inkog . --policy low-noise # only proven vulnerabilities inkog . --policy balanced # default — vulnerabilities + risk patterns inkog . --policy comprehensive # everything including hardening tips inkog . --policy governance # Article 14 controls, authorization, audit trails inkog . --policy eu-ai-act # EU AI Act compliance report ``` [Policy reference →](https://docs.inkog.io/cli/policies) </details> ## Community - 💬 [Discord](https://discord.gg/NuG4SSGRH) — questions, feedback, feature requests - 📚 [Documentation](https://docs.inkog.io) - 🐛 [Issues](https://github.com/inkog-io/inkog/issues) - 🤝 [Contributing](CONTRIBUTING.md) · [Changelog](CHANGELOG.md) <details><summary>Translations</summary> <a href="docs/i18n/README.zh-CN.md">简体中文</a> · <a href="docs/i18n/README.ja.md">日本語</a> · <a href="docs/i18n/README.ko.md">한국어</a> · <a href="docs/i18n/README.es.md">Español</a> · <a href="docs/i18n/README.pt-BR.md">Português</a> · <a href="docs/i18n/README.de.md">Deutsch</a> · <a href="docs/i18n/README.fr.md">Français</a> </details> ## License Apache 2.0 — see [LICENSE](LICENSE).