Abstract: The integration and deployment of machine learning (ML) models in configurable workflows is described. An example method of deployment of an ML model includes receiving a model submission package from a client device, extracting configuration parameters from a manifest file, provisioning a compute environment within the artificial intelligence (AI) platform based on the configuration parameters, instantiating and rendering an agent object within a graphical workflow canvas, and enabling, by the graphical workflow canvas, a user to define an executable workflow comprising the agent object, such that the executable workflow is invoked by the AI platform. The executable workflow can be invoked by the AI platform to perform model inference or service execution within the defined workflow.

Abstract: The system securely executes model context protocol (“MCP”) processes for artificial intelligence (“AI”) agents by creating nano sandboxes in which the MCP processes execute. A server-sent events (“SSE”) bridge identifies a request to initialize an SSE channel with an MCP server. The SSE bridge instantiates a nano sandbox based on a definition extracted from the request, creating a custom and dynamic isolated execution environment. An MCP process is launched within the nano sandbox. The SSE bridge returns a session identifier, receives a command and the session identifier, and forwards the command to the nano sandbox based on the session identifier. The MCP process executes the command within the nano sandbox, and responds over the SSE channel to the MCP client. When the SSE connection closes the nano sandbox can then be destroyed.

Abstract: Systems are disclosed for enabling secure cloud connections between artificial intelligence (“AI”) agents and client resources. A cloud connector executing at a client establishes an authenticated secure tunnel with a connection manager on a connection server. An AI agent is initiated to access cloud- and client-hosted AI models or data sources. The system uses guardrails to process inputs and outputs over the secure tunnel with the client, detecting prompt injection, confidential information, or personally identifiable information. Redundant secure tunnels and load balancing can be employed to ensure availability. Modified results are transmitted to a user device for display, providing AI agents with controlled access to client resources while enforcing confidentiality, integrity, and policy compliance.

Abstract: The invention provides a rules engine that applies customizable security rules and artificial intelligence (AI) model routing to queries. The rules are evaluated with respect to the query, an anticipate result of the query, and the actual result. Remedial actions are taken when the rule evaluation exceeds a threshold. The remedial actions include transforming the query by replacing sensitive information with a reversible placeholder. The actual result can be modified by reversing the placeholder and redacting additional sensitive information. The system can notify the user, an administrator, and a supervisor regarding the security evaluation and remedial actions. The evaluations can be logged for auditing purposes.

Abstract: Systems and methods are described for a messaging agent for managed multidimensional search based on an electronic message and management policies. A messaging application at a client device can send a new electronic message, such as an email, to a connector service. An attachment can be ingested and stored in a vector database. Then one or more artificial intelligence (“AI”) agents can be selected for responding to the body of the email, such as a query in the body. The responses can be formatted and sent to multiple users, such as a sending user of the electronic message and a recipient user that was copied or also sent the new electronic message. The AI agents can use different AI models, prompts, and vector databases depending on user permissions. This allows for building up vector databases with relevant content items and answering user questions based on those vector databases.

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 list of resource profiles, including tools and instructions, is received from the MCP server. This is stored and compared against a later list of tools and instructions. When a difference is detected, a user interface can display the difference to an administrator. Security rules cause the SSE bridge to be blocked from sending commands to the impacts tool or MCP server until approved by the administrator.

Abstract: A system detects changes in model context protocol (“MCP”) processes, and performs a remedial action. A user interface (“UI”) is generated that comprises a selections for identifying restricted actions, permissions for those restricted actions, triggers based on the restricted actions, and remedial actions that correspond to the triggers. The restricted actions can be based on a change to a resource profile of the MCP server. They can also apply to resource commands. The triggers can be specific keywords or change thresholds. The UI selections can be used to generate a management profile. Using the management profile, the system can automate remedial actions for MCP servers, protecting enterprises from data loss and other security issues.

Abstract: A gateway securely executes model context protocol (“MCP”) processes for artificial intelligence (“AI”) agents by creating nano sandboxes in which the MCP processes execute. A server-sent events (“SSE”) bridge identifies a request to initialize an SSE channel with an MCP server, the request being sent from an MCP client that generates MCP-compliant commands in association with an AI agent. The SSE bridge instantiates a nano sandbox based on a definition extracted from the request, creating a custom and dynamic isolated execution environment. An MCP process is launched within the nano sandbox. The SSE bridge returns a session identifier, and receives a command and the session identifier. The SSE bridge authorizes the command and response by applying security rules, blocking either for non-compliance.

Abstract: The present invention provides a system for processing user queries through an artificial intelligence (“AI”) pipeline, utilizing data chunking, question generation, and AI models to deliver contextually relevant responses. The system includes a server that ingests and chunks datasets, generates vector embeddings, and stores the data in one or more vector databases. A pipeline engine sends the chunked data to an AI model that generates potential questions tailored to different user personas. These questions, along with their corresponding data chunks, are stored in the database for future retrieval. When a user submits a query, the system semantically compares the query to the pre-generated question vectors and retrieves the most relevant question and associated data chunk. The query is then sent to an external AI model for final response generation. The system provides seamless interaction, delivering optimized, context-aware responses to user queries in real-time.

Abstract: A system allows users to utilize AI agents that access user-specific accounts (e.g., for tools) rather than just global tools that use a system login. The AI agent can execute multiple agent objects, such as an AI model. A tool can be assigned to an AI object, such that the tool is used when the AI object executes. An agent executor can check with an authentication service to determine whether the user is authenticated with a provider of the tool. If not, the agent executor can pause AI agent execution, store a current state, and prompt the user for authentication, which can appear in a window utilized by the AI agent. With a successful authentication, the agent executor resumes execution of the AI agent at the stored current state and stores an authorization token for use with the tool.

Abstract: The system provides a permissions engine derives permissions for responses from artificial intelligence (AI) agents. The permissions are derived based on the access permissions of chunks used to generate the response, and the response is cached for reuse with users meeting the permissions level. The system can also derive an access permission level for a user attempting to access the cached result or other content. This can be based on a managed access profile that includes user behavior criteria and device criteria. Likewise, the system can validate and derive sensitivity levels of resources that are ingested for use in the artificial intelligence (AI) agents and discriminate between chunks to use in synthesizing results based on the derived permissions of the chunks and their relationship to the user.

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.

Abstract: The invention provides a rules engine that manages user requests within an interface integrating multiple AI platforms and AI Models. Upon receiving a query, the engine assigns scores based on factors like management rules, user preferences, and contextual information. Determinative scores, such as those enforcing strict enterprise policies, can override others, leading the engine to block or reroute the query. If no score is determinative, the engine forwards the query and associated prompts to a specialized routing AI Model for contextual analysis. Based on this analysis, the rules engine directs the query to the most appropriate AI Model or defaults to the user-designated AI Model. This system balances user intent with rule enforcement, optimizing query processing across various AI platforms while ensuring compliance with enterprise guidelines.

Abstract: The invention provides a rules engine that manages user requests within an interface integrating multiple AI platforms and AI Models. Upon receiving a query, the engine assigns scores based on factors like management rules, user preferences, and contextual information. Determinative scores, such as those enforcing strict enterprise policies, can override others, leading the engine to block or reroute the query. If no score is determinative, the engine forwards the query and associated prompts to a specialized routing AI Model for contextual analysis. Based on this analysis, the rules engine directs the query to the most appropriate AI Model or defaults to the user-designated AI Model. This system balances user intent with rule enforcement, optimizing query processing across various AI platforms while ensuring compliance with enterprise guidelines.

Abstract: Retrieval of structured tabular data in a retrieval-augmented generation (RAG) platform is described. An example method includes receiving a tabular file including structured data, extracting short and long metadata from the file, and generating a dataset agent object using the extracted metadata. A human-readable description is associated with the structured data, and the dataset agent object and metadata are stored in a metadata index. A natural language query is received through a query agent object rendered within a canvas including interconnected agent objects. A dataset agent object is identified based at least in part on metadata filtering using the natural language query and short metadata. A structured query conforming to structured query language syntax is generated using a large language model (LLM) and executed on the structured data. A response including the result set and dataset agent identifier is generated and transmitted to a downstream agent object for further processing.

Abstract: Systems and methods are described for comparing execution of two or more artificial intelligence (AI) pipelines. A platform can provide a user interface (UI) that allows for selection or creation of multiple AI pipelines. The AI pipelines can utilize different pipeline objects, such as different prompts, datasets, or models. The AI pipelines can be displayed on a single UI screen, where execution of the AI pipelines is simultaneously simulated. The same inputs can be provided to the multiple AI pipelines, and the corresponding outputs can display on screen. The platform can also vectorize and compare the semantic similarity of the outputs, presenting an indicating of the semantic similarity on the same UI screen.

Abstract: Systems and methods are described for building artificial intelligence (“AI”) agents. A server can provide a user interface (UI) that includes options to select and connect various agent objects. The agent objects can include prompt objects, dataset objects, model objects, and one or more code objects. A subset of the agent objects can be identified as available for selection based on evaluation of at least one management policy associated with an administrative user. A manifest file is generated based on the connected agent objects, and the manifest is validated against dependency rules to ensure that the stages of the agent meet prerequisites for the stages. Then, the server performs a simulated execution of an agent that corresponds to the validated manifest file, including an identification of at least one execution metric associated with the simulated execution.

Abstract: Systems and methods are described for dynamic execution of artificial intelligence (“AI”) agents. A server can receive, from a client device, an input associated with an AI agent. Based on a manifest file or user profile, the server can identify a management policy that applies to the AI agent. The server then dynamically configures access to the agent objects based on applying the management policy. The management policy is applied to a device status of the client device, a user profile of a user of the client device, and/or a network configuration of the client device. The server then executes a modified workflow based on the dynamically configured access, wherein the modified workflow bypasses or changes operation of at least one of the agent objects. Based on the modified workflow, the server transmits an output to the client device.

Abstract: Systems and methods are described for tool discovery and ingestion for artificial intelligence (“AI”) agents. An AI platform can discover a first tool specification that includes an action and a description. The first tool specification is ingested to create a first tool object. The first tool object includes the action and an endpoint. Tool labels are determined from the specification and applied to the first tool object. A user interface displays the tool object, and it is added to an AI agent. The AI agent includes a manifest file that is used to execute the AI agent. This includes determining whether the AI agent is authorized to perform the action, and providing the AI agent with access to a tool credential, wherein the tool credential is sent to the endpoint.

Abstract: Systems and methods are described for a connected AI agent for managed multidimensional search based on an electronic message and management policies. A messaging application at a client device can send a new electronic message, such as an email, to a connector service. An attachment can be ingested and stored in a vector database. Then one or more artificial intelligence (“AI”) agents can be selected for responding to the body of the email, such as a query in the body. The responses can be formatted and sent to multiple parties, such as a sending user of the electronic message and a recipient that was copied or also sent the new electronic message. The AI agents can use different AI models, prompts, and vector databases depending on user permissions. This allows for building up vector databases with relevant content items and answering user questions based on those vector databases.