Enforce security policies on AI agent MCP calls
Security gateway that validates AI agent MCP calls against policies before execution to prevent harmful actions.
Why it matters
Protect users from poorly behaving AI agents by intercepting and validating every Model Context Protocol (MCP) call against defined security policies before allowing it to reach downstream servers.
Outcomes
What it gets done
Intercept all outbound MCP calls from AI agents before execution
Evaluate each call against configured security policy rules
Block or allow calls based on policy compliance checks
Provide a security gateway layer between agents and MCP servers
Install
Add it to your toolbox
Run in your project directory:
curl -fsSL https://spark.entire.vc/get/mcp-maybe-dont-ai-policy-engine | bash Overview
Maybe Don't AI Policy Engine
Maybe Don't AI is an MCP security gateway that provides policy enforcement for AI agent calls. It intercepts every call an agent makes to MCP servers and validates it against your defined policies before allowing it through. This protects your systems from agents that behave poorly or attempt unauthorized operations. Use Maybe Don't AI when you need to add security guardrails between AI agents and your MCP infrastructure, especially when agents have access to sensitive systems or can perform actions with real-world consequences. It's essential for teams running autonomous agents who want the benefits of agent capabilities while maintaining control over what operations are permitted.
What it does
Maybe Don't AI is a security gateway for the Model Context Protocol (MCP) that intercepts and validates every call an AI agent makes before it reaches downstream MCP servers. It enforces policy checks to protect users from agents that behave poorly or attempt unauthorized actions.
When to use - and when NOT to
Use Maybe Don't AI when you need to add a security layer between AI agents and your MCP servers, especially when agents have access to sensitive systems or data and you want to enforce rules about what operations they can perform. It's essential when you're running autonomous agents that could potentially make harmful or unintended calls to your infrastructure.
Do not use this tool if you're working in a completely sandboxed environment where agent actions have no real-world consequences, or if the overhead of policy evaluation would unacceptably slow down your agent's response time in latency-critical applications.
Inputs and outputs
The gateway receives MCP calls from AI agents as input, along with the policy rules you've configured to govern agent behavior. It outputs either approved calls that pass through to downstream MCP servers or blocked calls that violate your defined policies, protecting your systems from unauthorized or dangerous agent actions.
Who it's for
Maybe Don't AI is designed for developers and organizations deploying AI agents with access to production MCP servers who need to enforce security boundaries. It's particularly valuable for teams building autonomous agent systems where you want the benefits of agent capabilities but need guardrails to prevent misuse, errors, or security breaches. Unlike running agents with unrestricted access, this gateway provides a policy enforcement layer that sits between the agent and your infrastructure.
Source README
Yet another MCP security gateway, Maybe Don't AI provides policy checks on any call before it reaches downstream MCP servers to protect users from agents behaving poorly.
FAQ
Common questions
Discussion
Questions & comments · 0
Sign In Sign in to leave a comment.