# Maybe Don't, AI > AI Guardrails - Real-time monitoring and intelligent blocking for AI agents Maybe Don't provides guardrails and observability for agentic AI. It sits between AI agents and the tools they use — MCP servers and CLI commands — evaluating every operation against your policies before it executes. You define the rules, Maybe Don't enforces them. ## Getting Started - [Get Started](https://maybedont.ai/docs/get-started/): Install, configure, and run Maybe Don't - [Configuration Reference](https://maybedont.ai/docs/configuration/reference/): Complete reference for maybe-dont.yaml ## MCP Gateway - [MCP Gateway Overview](https://maybedont.ai/docs/mcp-gateway/): How the MCP proxy works - [Downstream Servers](https://maybedont.ai/docs/mcp-gateway/downstream-servers/): Configure MCP servers to proxy - [Server Configuration](https://maybedont.ai/docs/mcp-gateway/server-modes/): Transport modes, listen address, and trusted proxies ## CLI Gateway - [CLI Gateway Overview](https://maybedont.ai/docs/cli-gateway/): Validate CLI commands from AI agents - [Validate API](https://maybedont.ai/docs/cli-gateway/rest-api/): REST endpoint for CLI command validation ## Policies - [AI Policies](https://maybedont.ai/docs/policies/ai-policies/): Natural language policies evaluated by an LLM - [CEL Policies](https://maybedont.ai/docs/policies/cel-policies/): Deterministic policies using Common Expression Language - [How-To](https://maybedont.ai/docs/policies/writing-policies/): Practical guide for writing policies ## Audit Log - [Audit Log](https://maybedont.ai/docs/audit-log/): The record of everything - [Log Schema](https://maybedont.ai/docs/audit-log/log-schema/): Complete field reference for audit log entries - [SIEM](https://maybedont.ai/docs/audit-log/siem-integration/): Ingesting audit logs into your SIEM ## Testing - [Test Cases](https://maybedont.ai/docs/testing/test-cases/): Define operations and expected policy decisions - [Suites](https://maybedont.ai/docs/testing/test-suites/): Directory layout and suite.yaml configuration - [Runner](https://maybedont.ai/docs/testing/test-runner/): CLI commands, output formats, and CI/CD integration ## Configuration - [Logging](https://maybedont.ai/docs/configuration/logging/): Application and audit log configuration ## Agent Integrations - [Agents Overview](https://maybedont.ai/docs/agents/): Choosing between MCP gateway, hooks, and skills ### MCP Gateway Connections - [Claude Code](https://maybedont.ai/docs/agents/mcp/claude-code/): Connect Anthropic's CLI assistant via MCP - [GitHub Copilot](https://maybedont.ai/docs/agents/mcp/github-copilot/): Connect GitHub's AI assistant via MCP - [Cursor](https://maybedont.ai/docs/agents/mcp/cursor/): Connect the AI-powered code editor via MCP - [OpenAI Codex](https://maybedont.ai/docs/agents/mcp/openai-codex/): Connect OpenAI's coding CLI via MCP - [Gemini Code Assist](https://maybedont.ai/docs/agents/mcp/gemini-code-assist/): Connect Google's AI assistant via MCP - [OpenHands](https://maybedont.ai/docs/agents/mcp/openhands/): Connect the open-source AI developer via MCP - [Cody](https://maybedont.ai/docs/agents/mcp/cody/): Connect Sourcegraph's coding assistant via MCP ### Hook Integrations - [Hooks Overview](https://maybedont.ai/docs/agents/hooks/): Integrate agents via hook scripts and the intercept endpoint - [Claude Code Hooks](https://maybedont.ai/docs/agents/hooks/claude-code/): Claude Code PreToolUse/PostToolUse hooks - [Cursor Hooks](https://maybedont.ai/docs/agents/hooks/cursor/): Cursor shell and MCP execution hooks - [Gemini CLI Hooks](https://maybedont.ai/docs/agents/hooks/gemini-cli/): Gemini CLI BeforeTool/AfterTool hooks - [Cline Hooks](https://maybedont.ai/docs/agents/hooks/cline/): Cline preToolUse/postToolUse hooks - [GitHub Copilot Hooks](https://maybedont.ai/docs/agents/hooks/github-copilot/): GitHub Copilot PreToolUse/PostToolUse hooks ## Solutions - [AI Guardrails](https://maybedont.ai/solutions/ai-guardrails/): Runtime policy enforcement for AI agents - [Agentic Observability](https://maybedont.ai/solutions/agentic-observability/): Complete visibility into every MCP tool call and CLI command your AI agents execute - [Governance & Compliance](https://maybedont.ai/solutions/governance-compliance/): Meet ISO 42001, SOX, HIPAA, and FINRA requirements for AI agent operations ## Other - [Skills](https://maybedont.ai/docs/agents/skills/): Pre-written instructions that teach AI agents how to use Maybe Don't - [Native Tools](https://maybedont.ai/docs/native-tools/): Experimental MCP tools built into Maybe Don't - [Security](https://maybedont.ai/docs/security/): Security model and deployment considerations