# MintMCP MCP for AI Agents AI Agent Connect > MintMCP is an enterprise-grade gateway that gives you control over how your AI agents interact with internal data. It handles security guardrails, PII redaction, and RBAC policies through a centralized proxy system. You can audit every action, manage virtual servers, and revoke access instantly to keep your organization compliant with SOC2 standards. ## Overview - **Category:** fort-knox - **Price:** Free - **Endpoint:** https://edge.vinkius.com/vk_preview_O150J9v0kxg8rdyosRWdbRwIt3xDeHzZmIR4xy76/ai-agent-connect - **Tags:** governance, security-guardrails, pii-redaction, soc2, proxy-routing ## Description Imagine you're trying to let your AI agents touch company data without worrying about them leaking private info or hitting things they shouldn't see. MintMCP acts as the gatekeeper. It creates a layer of protection where every request from your agent gets checked against your specific security rules before it ever hits your internal systems. You can set up virtual servers to group tools by function, making it easier to manage who gets access to what. If something goes wrong or a connection gets compromised, you can kill it immediately without affecting the rest of your setup. It's built for teams that need to move fast but can't afford to skip the compliance steps. By using this Connector, you get a clear audit trail of every single action the agent takes, which makes staying SOC2 compliant a lot less of a headache. It's a core piece of infrastructure for anyone who needs to move from experimenting with AI to running AI in production without the security team breathing down your neck. You'll find it's one of the most reliable ways to manage these connections within the Vinkius catalog. ## Tools ### mintmcp_get_tool_policy Fetches the specific RBAC and governance parameters for a tool. You can check exactly what rules are in place for a specific function. ### mintmcp_get_virtual_server Pulls the exact configuration for a single Virtual Server. Use it to inspect the schema of a specific proxy group. ### mintmcp_list_available_tools Shows which tools are currently approved inside a specific Virtual Server. This helps you see exactly what your agent can actually do right now. ### mintmcp_list_virtual_servers Lists all the proxy abstractions you've set up to group tools. Use this to see your entire organized topology at a glance. ### mintmcp_revoke_access_token Ends an active OAuth flow immediately. It's your panic button to isolate a compromised session. ### mintmcp_run_tool_action Proxies an execution stream to your native integrations. This is how the agent actually performs work through the gateway. ### mintmcp_eval_guardrail Checks a string against your security and PII redaction engines. It ensures no sensitive data slips through the cracks. ### mintmcp_fetch_audit_logs Dumps the systematic telemetries for your SOC2 matrix. It provides a complete history of every access event. ## Prompt Examples **Prompt:** ``` Can you show me which tools are available in our 'Finance' proxy? ``` **Response:** ``` | Proxy: Finance | Available Tools: | | `mintmcp_list_available_tools` | [Internal_Ledger_Read] | [Expense_Report_Submit] | [Currency_Converter] | *Note: These are the only tools authorized for the Finance virtual server.* ``` **Prompt:** ``` Check if this message contains any private info: 'My name is John Doe and my account number is 12345'. ``` **Response:** ``` I've run that through the security guardrails. **Result: Blocked.** The system identified a potential account number and redacted it. The agent will not be able to send this message to the external system. ``` **Prompt:** ``` Give me a summary of the last 5 actions taken by the HR agent. ``` **Response:** ``` Here is the audit summary for the HR agent: * **Action 1:** Viewed employee handbook (Success) * **Action 2:** Updated vacation request (Success) * **Action 3:** Viewed payroll stats (Denied - Insufficient Permissions) * **Action 4:** Updated vacation request (Success) * **Action 5:** Viewed employee handbook (Success) ``` ## Capabilities ### Block PII leaks Automatically redacts sensitive data before it leaves your environment. ### Group tools by function Organizes different internal plugins into virtual server proxies. ### Audit every action Logs every single request for security reviews and compliance reporting. ### Manage permissions Sets rigid rules on which functions specific users or agents can run. ### Kill compromised sessions Revokes access tokens instantly to isolate security risks. ## Use Cases ### Preventing PII leaks in customer support A security team needs to ensure an agent doesn't leak customer emails. They use mintmcp_eval_guardrail to scan every prompt and response for PII. ### Departmental tool provisioning An admin needs to see what tools are available to a specific department. They call mintmcp_list_available_tools to show the approved set for that department's virtual server. ### Post-incident security auditing A developer needs to audit a series of automated actions from last night. They use mintmcp_fetch_audit_logs to pull the full telemetry of all executed runs. ### Emergency session termination A compromised API key is detected. The admin uses mintmcp_revoke_access_token to kill the session immediately and protect the host. ## Benefits - Stop PII leaks before they happen. Use mintmcp_eval_guardrail to ensure your agent never sees or sends sensitive customer data. - Get a clear audit trail. mintmcp_fetch_audit_logs gives you every single action in a format that satisfies your security team. - Manage access with precision. Use mintmcp_get_tool_policy to define exactly who can run which function based on your RBAC rules. - Isolate risks instantly. If an agent behaves weirdly, mintmcp_revoke_access_token cuts it off from your system in seconds. - Organize your internal tools. mintmcp_list_virtual_servers lets you group plugins into logical proxies rather than a messy list. ## How It Works The bottom line is that it puts a secure, auditable wall between your AI agents and your sensitive company data. 1. Get your MintMCP variables ready, including your Tenant ID and primary API Token. 2. Set your context engine limits to define exactly what your agent is allowed to do. 3. Request logs and run telemetry to see exactly how your agent is interacting with the proxy. ## Frequently Asked Questions **How does MintMCP help with SOC2 compliance?** It provides a centralized audit trail for every action your AI agent takes. By logging all requests and enforcing RBAC policies, it makes it much easier to prove to auditors that your AI workflows are secure and governed. **Can MintMCP block my AI agent from sharing sensitive data?** Yes, that's a core feature. It uses built-in guardrails to scan for PII and other sensitive information, redacting it before it ever leaves your secure environment. **What are virtual servers in MintMCP?** Think of them as organized folders for your tools. Instead of giving an agent access to everything, you group specific tools into virtual servers and only give the agent access to the ones it actually needs. **Can I quickly cut off an AI agent if it starts acting up?** Absolutely. You can instantly revoke access tokens to kill a session. This isolates any potential issues without having to shut down your entire system. **Is MintMCP suitable for large enterprise environments?** It's specifically designed for them. It's built to handle complex governance, multi-tenant identities, and the kind of rigid security requirements that large companies face when deploying AI. **Does MintMCP work with my existing internal tools?** Yes, it acts as a proxy gateway. It sits between your AI client and your existing internal systems, routing the logic safely while applying your security rules. **Can I test payload parameters through the security Guardrails offline?** Yes, mapping direct values through `eval_guardrail` runs simulated queries bypassing the actual downstream proxy directly into the evaluation loop. **Are explicit SOC2 Audit Logs available natively to query via this agent integration?** Yes! Utilize `fetch_audit_logs` limiting explicitly standard parameters to unpack all tool executions securely bounded inside your tenant matrix natively. **How do virtual servers separate functional tools proxy instances natively?** By polling `list_available_tools` against a specific server ID, you extract rigidly grouped schemas bounded solely to that environment parameter.