BlueNexus is a central player in the agentic infrastructure stack, providing the hosting environment, memory layer, and tool-access protocols necessary for LLMs to function as autonomous agents. They are particularly relevant for their support of the Model Context Protocol (MCP), which facilitates standardized communication between agents and external data sources.
By offering hardware-isolated compute (TEEs) and a marketplace for domain-specific agents, they address the two biggest hurdles in the current agent ecosystem: security and distribution. They matter to the ecosystem because they are championing "provable privacy," moving the industry away from simple policy-based trust toward technical guarantees. This makes them a key partner for builders creating agents for sensitive or regulated use cases.
BlueNexus is an Australian technology company building the infrastructure required to turn Large Language Models into persistent, task-oriented agents. Founded by Nicholas Arbuckle and based in Surry Hills, Sydney, the company focuses on the gap between conversational AI and functional digital assistants that can operate independently and securely.
While most consumer AI experiences are ephemeral, occurring within a single chat transcript, BlueNexus structures its output as "Pages." These are versioned, editable documents where agents and users can collaborate synchronously. The intent is to move the user experience from a text-in-text-out queue to a shared workspace. This architecture supports memory that compounds over time, building a knowledge base from past conversations and user uploads rather than relying on the context window of a single session.
The core technical differentiator for BlueNexus is its use of Trusted Execution Environments (TEEs). Users can toggle Standard-TEE mode, which runs agent compute inside hardware-isolated environments. This ensures that the data being processed is invisible to the platform operators. This feature is a direct appeal to regulated industries like legal and finance where data leakage is a non-starter. By enforcing privacy through code and hardware rather than just policy, BlueNexus attempts to solve the trust deficit that currently prevents broader AI adoption in enterprise settings.
The platform includes over 100 native connectors across categories like CRMs, development tools, and home devices. Crucially, BlueNexus supports the Model Context Protocol (MCP), allowing users to connect any MCP-compatible server to their agents. This open approach to tools prevents vendor lock-in and allows the agents to act on a schedule, such as drafting morning briefs or processing data queues at midnight. For developers and experts, BlueNexus operates a marketplace where they can publish and monetize their specialized agents. This allows doctors, lawyers, or strategists to package their specific domain expertise into a usable agent that others can hire.
BlueNexus utilizes a usage-based pricing model that avoids traditional monthly subscriptions. Users pay for "pages" of agent compute, where one page is defined as 3,000 characters. This covers the total work performed by the agent, including messages, reasoning, and tool results. The billing is proportional to the work performed; short queries cost very little, while long-running multi-turn agents consume credits as the conversation context grows. New users typically receive a credit grant to test the system without a credit card. This model aligns the company's revenue with the actual computational costs of the agents, reflecting a shift away from the "data for access" trade common in the first wave of consumer AI.
A platform for building and deploying privacy-first AI agents with long-term memory and app integrations.
BlueNexus is hiring.