Paradigm Leverages Centaur AI Agent Runtime for Internal Operations

Paradigm has announced the open-sourcing of Centaur, an AI Agent Runtime designed to support secure execution and multi-user collaboration. The tool, which was jointly developed by Paradigm and Tempo, is a self-hosting runtime that allows for the deployment of autonomous AI agents within an organization’s own infrastructure.

According to reporting from Odaily, Centaur functions as a shared agent capable of executing tasks continuously over periods ranging from several hours to multiple days. Unlike standard AI chatbots that operate on a prompt-and-response basis, Centaur can maintain its state and continue tasks even after a system restart.

The runtime is engineered to handle sensitive operations by utilizing actual credentials to complete tasks without ever coming into direct contact with raw secrets. This security architecture is intended to allow agents to interact with various tools and systems while minimizing the risk of credential exposure.

Users can interact with Centaur and trigger its capabilities through two primary interfaces: an API or via Slack.

Paradigm began utilizing Centaur internally in January 2026. The fund stated that the implementation of the agent runtime has transformed its internal operational processes across several different business functions.

The internal application of Centaur has spanned the following areas:

• Investment analysis and operations
• Engineering and design workflows
• Recruitment and hiring processes
• Event planning and management
• Customer service operations

The transition to an open-source model for Centaur allows other developers and organizations to implement a similar shared-agent architecture. By providing a runtime that supports long-term execution and secure credential management, the project addresses common bottlenecks in AI agent deployment, such as session timeouts and the security risks associated with granting AI models access to private API keys or passwords.