# Cynative

> Ask your infrastructure anything. An open-source framework for building security agents with live, read-only access to your code, cloud and Kubernetes: an AI-native cybersecurity platform in one binary, Apache-2.0.
Cynative runs frontier models across GitHub, GitLab, AWS, GCP, Azure and Kubernetes as one system and returns verified answers. It writes and runs code in an ephemeral sandbox and queries APIs in parallel, and one question fans out across the whole stack. Each finding is cross-checked and traced to its origin.
Read-only by construction. Each call is gated and authorized before a credential is attached. Cloud native security software in one static binary, written in pure Go, that runs entirely in your environment and keeps every result there.

- [Cynative: AI-Powered Cybersecurity Platorm](https://docs.cynative.com/index.md): Open-source CLI framework that runs frontier LLMs over AWS, GCP, Azure, Kubernetes, GitHub, and GitLab to perform read-only security deep research.
- [Cynative: AI Security Research for Cloud Infrastructure](https://docs.cynative.com/introduction.md): Cynative runs frontier AI models over AWS, GCP, Azure, Kubernetes, and GitHub to deliver read-only cloud security research with verified findings.
- [Cynative Quickstart: Install, Configure, and Run in Minutes](https://docs.cynative.com/quickstart.md): Install Cynative, configure an LLM provider, and run your first cloud security audit with a built-in agent or a natural-language question.
- [How to Install Cynative on macOS, Linux, and Windows](https://docs.cynative.com/installation.md): Install the Cynative CLI via Homebrew, install script, Scoop, or manual download. Verify release signatures with Sigstore and Cosign for integrity.
- [Running Agents: Interactive, Seeded, and Print Modes](https://docs.cynative.com/running-agents.md): Run built-in agents or free-form questions interactively, as seeded tasks, or non-interactively in CI pipelines with exit-code-driven scripting.
- [Interactive Sessions, Approvals, and Stopping Cynative](https://docs.cynative.com/sessions-and-approvals.md): Understand Cynative's interactive session model, the per-tool-call approval flow, how to stop a run gracefully, and how to use auto-approve for CI.
- [Cynative JavaScript Sandbox and Concurrent Code Execution](https://docs.cynative.com/code-execution.md): Cynative's JS sandbox lets the model write concurrent API scripts using http_request to fan out research across cloud resources efficiently.
- [Audit Log: Fail-Closed JSONL Record of Every Tool Call](https://docs.cynative.com/audit-log.md): Cynative's JSONL audit log records every tool call with agent provenance, decisions, and redacted results. Configure path, rotation, and retention.
- [Cynative Agents: Named Prompts for Security Research](https://docs.cynative.com/agents/overview.md): Agents are Markdown files that supply the prompt for a Cynative run. Use built-in agents, shadow them, or write your own for repeatable security research.
- [Built-in Agent Catalog: All 45 Cloud Security Agents](https://docs.cynative.com/agents/catalog.md): All 45 built-in Cynative agents for AWS, Azure, GCP, GitHub, and Kubernetes — with one-line descriptions and run commands for each.
- [Writing Custom Cynative Agents: Format and Naming Rules](https://docs.cynative.com/agents/writing-agents.md): Create a custom Cynative agent in one Markdown file. Covers the frontmatter schema, naming rules, shadowing behavior, and piped input handling.
- [Cynative Connectors: Multi-Cloud Credential Discovery](https://docs.cynative.com/connectors/overview.md): Cynative connectors discover credentials from your shell environment and enforce read-only access to AWS, GCP, Azure, Kubernetes, GitHub, and GitLab.
- [AWS Connector: Credential Discovery and Read-Only Hardening](https://docs.cynative.com/connectors/aws.md): Configure Cynative's AWS connector: credential chain, STS credential downscoping to SecurityAudit, region selection, and EKS cluster access.
- [GCP Connector: Application Default Credentials Access](https://docs.cynative.com/connectors/gcp.md): Configure Cynative's GCP connector: Application Default Credentials, roles/viewer action gate, project selection, and GKE cluster access.
- [Azure Connector: Reader Role and ARM Credential Enforcement](https://docs.cynative.com/connectors/azure.md): Configure Cynative's Azure connector: Azure credential chain, Reader role enforcement, subscription selection, cloud environment, and AKS cluster access.
- [Kubernetes Connector: Self-Managed Cluster Security Research](https://docs.cynative.com/connectors/kubernetes.md): Configure Cynative's self-managed Kubernetes connector: kubeconfig discovery, live view RBAC enforcement, custom ClusterRole, and API server access.
- [GitHub and GitLab Connectors: Source Control Research](https://docs.cynative.com/connectors/github-gitlab.md): Configure Cynative's GitHub and GitLab connectors: token discovery, read-only enforcement, permission ceilings, and GitLab self-managed host configuration.
- [LLM Provider Overview: 30 Backends Supported by Cynative](https://docs.cynative.com/providers/overview.md): Cynative supports 30+ LLM providers via the Bifrost SDK. Configure your provider with two environment variables or a YAML file for advanced options.
- [Configure OpenAI Models as Cynative's LLM Provider](https://docs.cynative.com/providers/openai.md): Set up Cynative with OpenAI's GPT and o-series models. Configure API key, model selection, custom base URL for proxies, and advanced network options.
- [Configure Anthropic Claude as Cynative's LLM Provider](https://docs.cynative.com/providers/anthropic.md): Set up Cynative with Anthropic's Claude models. Configure API key, model selection, reasoning effort, and prompt caching for cost efficiency.
- [Configure Amazon Bedrock as Cynative's LLM Provider](https://docs.cynative.com/providers/bedrock.md): Run Cynative with AWS Bedrock-hosted models. Uses the AWS credential chain — no API key needed. Configure region, model ID, and cross-region inference.
- [Vertex AI and Gemini Provider Configuration in Cynative](https://docs.cynative.com/providers/vertex.md): Run Cynative with Google Vertex AI or Gemini models. Uses Application Default Credentials. Configure project ID, region, and model selection.
- [Local Inference with Ollama, vLLM, or SGLang in Cynative](https://docs.cynative.com/providers/local.md): Run Cynative against a local or self-hosted model via Ollama, vLLM, or SGLang. No API key required — point Cynative at your endpoint and pick a model.
- [Cynative Configuration: YAML File and Environment Variables](https://docs.cynative.com/configuration.md): Complete reference for ~/.cynative/config.yaml and every CYNATIVE_* env var — LLM provider, connectors, audit log, cache, and session limits.
- [Cynative Outbound Proxy: HTTPS_PROXY and NO_PROXY Setup](https://docs.cynative.com/proxy.md): Configure HTTPS_PROXY, HTTP_PROXY, and NO_PROXY for Cynative — what traffic routes through the proxy and what an intercepting proxy can observe.
- [Cynative Architecture: Single Binary, No External State](https://docs.cynative.com/project/architecture.md): Cynative is a single static Go binary. Understand the agent loop, tool system, action gate, JS sandbox, connectors, and audit pipeline.
- [Cynative Roadmap: Direction, Priorities, and Non-Goals](https://docs.cynative.com/project/roadmap.md): Cynative's planned direction: broader connector coverage, MCP support, GitHub Actions integration, agent memory, and what we explicitly will not build.
- [Cynative Security Model: Trust Boundaries and Controls](https://docs.cynative.com/security.md): Cynative's defense-in-depth security model: action gate, host pinning, JS sandbox, credential handling, and how to report vulnerabilities.
- [Contributing to Cynative: Setup, Gates, and PR Flow](https://docs.cynative.com/project/contributing.md): Set up your Cynative development environment, understand the CI gates, write tests, and follow PR conventions for squash-merge linear history.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.