IP Library Granted Patent US 12,683,991
Granted Patent B1
US 12,683,991 · App. 19/529,562 · Granted Jul 14, 2026

Insecure model context protocol server remediation for artificial intelligence agents

Inventors: Andrew Morgan (Dublin, IE); Rahul Parwani (Atlanta, GA)
Assignee: Airia LLC
H04L63/1416G06F21/53H04L41/40H04L63/102H04L63/105H04L63/1441H04L63/20H04L67/02
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,683,991
App. No.
19/529,562
Filed
Feb 4, 2026
Granted
Jul 14, 2026
Kind
B1
Art Unit
2435
USPC
726/1
Abstract

A system detects changes in model context protocol (“MCP”) processes, and performs a remedial action. A server-sent events (“SSE”) bridge sends a request to an MCP server. A first resource profile is received from the MCP server. This is stored and compared against a second updated version of the resource profile. When a difference is detected, the SSE bridge determines whether to block a resource command from reaching the SSE bridge. The decision is based on comparing the difference to security rules, which can be defined as part of a management profile.

Claims (37)

1 . A method for securely executing artificial intelligence (“AI”) agents by dynamically reacting to changes in model context protocol (“MCP”) tools, comprising:

monitoring, by an MCP orchestrator, a resource available at an MCP server, the resource being defined by a resource profile that specifies one or more invocation parameters of the resource;

detecting a modification to the resource profile;

receiving a resource command directed to the resource, the resource command being formed in accordance with the resource profile;

automatically determining a remedial action based on the detected modification; and

applying the remedial action to a communication between the MCP orchestrator and the MCP server, the communication comprising at least one of: (i) the resource command, or (ii) a receipt of a response from the MCP server.

2 . The method of claim 1 , wherein the modification comprises a new command signature that does not exist in a prior version of the resource profile.

3 . The method of claim 1 , wherein the modification comprises a change to an instruction associated with the resource, wherein the instruction is provided to an AI model to guide invocation of the resource.

4 . The method of claim 1 , wherein the modification comprises a change to an endpoint identifier associated with the resource.

5 . The method of claim 1 , wherein detecting the modification comprises comparing a first version of the resource profile against a second version of the resource profile.

6 . The method of claim 1 , further comprising generating a deviation score based on the detected modification, wherein the remedial action is determined based on comparing the deviation score against at least one threshold.

7 . The method of claim 6 , wherein the deviation score is generated by an AI model based on semantic analysis of the modification.

8 . The method of claim 1 , wherein the communication comprises a transmission of the resource command to the MCP server.

9 . The method of claim 8 , wherein the remedial action comprises blocking the transmission of the resource command to the MCP server.

10 . The method of claim 8 , wherein the remedial action comprises permitting the transmission of the resource command to a first version of the MCP server while blocking transmission to a second version of the MCP server.

11 . The method of claim 8 , wherein applying the remedial action comprises validating the resource command against security rules prior to the transmission to the MCP server.

12 . The method of claim 1 , wherein the communication comprises the receipt of the response from the MCP server.

13 . The method of claim 12 , wherein applying the remedial action to the receipt of the response comprises validating the response against security rules prior to transmitting the response to an MCP client.

14 . The method of claim 1 , wherein the remedial action comprises transmitting an alert to an administrative user.

15 . The method of claim 1 , wherein the remedial action comprises modifying permissions associated with the resource command.

16 . The method of claim 1 , wherein the remedial action comprises terminating the MCP server and deleting a nano sandbox in which the MCP server executed.

17 . The method of claim 1 , further comprising instantiating a nano sandbox for execution of the MCP server, wherein the nano sandbox comprises an isolated execution environment.

18 . The method of claim 17 , wherein the remedial action comprises limiting system resources provided within the nano sandbox.

19 . A non-transitory, computer-readable medium having instructions for securely executing artificial intelligence (“AI”) agents by dynamically reacting to changes in model context protocol (“MCP”) tools, wherein the instructions are executed by a processor, causing the processor to perform stages comprising:

monitoring, by an MCP orchestrator, a resource available at an MCP server, the resource being defined by a resource profile that specifies one or more invocation parameters of the resource;

detecting a modification to the resource profile;

receiving a resource command directed to the resource, the resource command being formed in accordance with the resource profile;

automatically determining a remedial action based on the detected modification; and

applying the remedial action to a communication between the MCP orchestrator and the MCP server, the communication comprising at least one of: (i) the resource command, or (ii) a receipt of a response from the MCP server.

20 . A system for securely executing artificial intelligence (“AI”) agents by dynamically reacting to changes in model context protocol (“MCP”) tools, comprising:

a memory storage including a non-transitory, computer-readable medium comprising instructions; and

at least one hardware-based processor that executes the instructions to carry out stages comprising:

monitoring, by an MCP orchestrator, a resource available at an MCP server, the resource being defined by a resource profile that specifies one or more invocation parameters of the resource;

detecting a modification to the resource profile;

receiving a resource command directed to the resource, the resource command being formed in accordance with the resource profile;

automatically determining a remedial action based on the detected modification; and

applying the remedial action to a communication between the MCP orchestrator and the MCP server, the communication comprising at least one of: (i) the resource command, or (ii) a receipt of a response from the MCP server.