Secure tunnel access to remotely executed tools for artificial intelligence agents

A non-transitory, computer-readable medium containing instructions for artificial intelligence (“AI”) agents to utilize client-hosted tools over secure cloud connections, wherein the instructions are executed by at least one processor to perform stages comprising:

• receiving, from a cloud connector executing on a client, an authentication request at a connection server, wherein the connection server is instantiated by an AI platform;
• opening a first secure tunnel between the client and the connection server;
• initiating execution of an AI agent, wherein the AI agent is configured to access a model context protocol (“MCP”) server that executes at the client;
• transmitting a first resource command over the secure tunnel, the resource command being associated with the MCP server;
• receiving an output of the MCP server;
• modifying the output with guardrails, wherein the guardrails are configured to detect at least one of prompt injection, confidential information, and personally identifiable information (“PII”);
• and
• utilizing the output or a modified output in executing an agent object of the AI agent, wherein the agent object comprises one of an AI model, a vector database, a dataset, and a script.

A system for artificial intelligence (“AI”) agents to utilize client-hosted tools over secure cloud connections, the system comprising:

• at least one physical non-transitory, computer-readable medium comprising instructions;
• and
• at least one processor that executes the instructions to perform stages comprising:
• receiving, from a cloud connector executing on a client, an authentication request at a connection server, wherein the connection server is instantiated by an AI platform;
• opening a first secure tunnel between the client and the connection server,
• initiating execution of an AI agent, wherein the AI agent is configured to access a model context protocol (“MCP”) server that executes at the client;
• transmitting a first resource command over the secure tunnel, the resource command being associated with the MCP server;
• receiving an output of the MCP server;
• modifying the output with guardrails, wherein the guardrails are configured to detect at least one of prompt injection, confidential information, and personally identifiable information (“PII”);
• and
• utilizing the modified output in executing an agent object of the AI agent, wherein the agent object comprises one of an AI model, a vector database, a dataset, and a script.