Secure, Addressable, and Context-Aware AI Agents with atPlatform™

Ensure AI agents are secure, addressable, and able to participate in context-aware exchanges.

As AI grows from general LLMs into more focused, expert models, we must make sure AI agents can operate as trusted entities within complex and evolving environments.

Here’s how the atPlatform achieves this:

Addressability and Identity

AI agents are assigned unique identifiers, called atSigns, that also function as an address used for peer-to-peer communication on the atPlatform. Each one is unique and cannot be impersonated or forged.

Authentication and Secure Data Exchange

The atPlatform confirms identity through Zero Trust cryptography before allowing a connection to be made. Once connected, AI agents can exchange end-to-end encrypted data, which can only be decrypted by the intended recipient.

Contextual Awareness

Each AI agent’s unique atSign enables private, accurate, and relevant responses based on the identity of the requester. The atPlatform supports Model Context Protocol enabled applications.

Policy Plane

Rulesets can be predefined, associated with individuals or customizable groups, and assigned consistently across AI agents, models, data sources, and the people that need to access them. Policy is dynamic, and access can be assigned for a single session, an hour, or something more granular.

Invisibility

All communication between people, entities, and things is done without requiring any open listening ports at either end. This means when someone scans the network, the AI agents cannot be detected even though they’re online, eliminating their network attack surface.

Direct API Communication

atPlatform allows AI agents to directly communicate with APIs that are also atPlatform-enabled, creating a seamless and secure ecosystem.

Protecting Model Context Protocol (MCP) Servers

MCP servers act as API interfaces for AI interaction, so their secure deployment is crucial for protecting proprietary data and intellectual property.

With the atPlatform, it’s possible to create an end-to-end encrypted solution for securing access to MCP servers from the internet, directly addressing the inherent vulnerabilities associated with exposing these interfaces. These vulnerabilities include: Denial-of-Service (DoS) attacks, brute-force attacks against authentication methods, and the risk of leaked API keys. Developers and IT professionals can deploy and manage MCP servers with enhanced security, ensuring the privacy and integrity of their AI operations.

Key Technical Advantages for MCP Server Security:

Eliminates Open Ports – By removing the need for open ports, MCP servers become invisible to external port scanners and other reconnaissance techniques. This makes the servers undetectable to unauthorized entities and removes the network attack surface.
Uses End-to-End Encryption – All data exchanged with MCP servers is protected by end-to-end encryption. This ensures the confidentiality and integrity of sensitive data, proprietary logic, and AI interactions from the point of origin to the intended recipient, preventing eavesdropping and tampering.
Secures API Exposure for LLMs – This solution allows organizations to expose their MCP server APIs to LLMs without direct public internet exposure. This enables secure integration of proprietary AI models and internal data with LLM capabilities, facilitating context-sensitive AI responses based on internal policies and data.

The atPlatform provides structure and control over the dynamic landscape of AI, enabling practical, manageable, and secure implementations.

Here’s a scenario featuring AI agents that illustrates why privacy, addressability, and context are necessary:

Sarah, the VP of Global Supply Chain at a large manufacturing company, is managing a critical production run. Her AI assistant, connected to the Internet via the atPlatform, messages her:

“Sarah, we’ve detected a potential supply chain disruption impacting the Q4 production forecast for our flagship product. A key component from Supplier X is experiencing a 3-day delay.”

Sarah responds, “That’s concerning. Can we identify alternative suppliers and assess the impact on our delivery timelines?“

“Analyzing alternative options now,”

her AI replies. In the background, her AI securely communicates with the AI agents of various suppliers to gather real-time inventory and delivery data. It also interfaces with the company’s internal ERP and logistics systems.

After processing the data, Sarah’s AI reports,

“Supplier Y can provide the component with a 2-day lead time, minimizing the impact. However, their pricing is 15% higher. I’ve also checked with the logistics AI to optimize the delivery schedule, minimizing delay to 1 day if we utilize Supplier Y.”

Sarah then asks, “Can we negotiate with Supplier Y for a better price, and also check if legal has a pre-existing contract with them?“

Her AI replies,

“Initiating negotiation with Supplier Y’s AI now, while simultaneously querying the legal contract database. Also, the logistics AI is adjusting the delivery schedule to account for potential delays.”

Moments later, Sarah’s AI updates her,

“Negotiations with Supplier Y have resulted in a 10% price decrease. Legal has confirmed an existing framework agreement with Supplier Y, simplifying the procurement process. The logistics AI has finalized the optimized delivery schedule, ensuring minimal disruption. Do you want to approve the purchase order and notify the relevant stakeholders?”

Sarah replies, “Yes, proceed and notify the production and sales teams.“

Her AI confirms,

“Purchase order approved, notifications sent, and the production schedule has been automatically updated. All actions are logged and auditable for compliance.”