Agentao Offers Governed Runtime for Tool-Using LLM Agents

As Large Language Model (LLM) agents become more sophisticated, interacting with tools, managing local state, and using persistent memory, new security and governance challenges emerge. Risks such as over-privileged actions, weak auditability, prompt injection, and uncontrolled side effects are significant concerns for developers and deployers. Agentao is introduced as a solution: a governed, local-first runtime specifically built for these advanced, tool-using LLM agents. Its core design principle is to explicitly separate the actions proposed by the LLM from the execution authorized by the host system. This is achieved through a layered architecture that includes host-facing surfaces, a host contract, a runtime core, and a permission-mediated tool system, alongside subsystems for memory, replay, and protocol integration. While Agentao does not provide formal safety guarantees, it significantly enhances the governability, inspectability, and suitability of agents for host-controlled local environments by making permissions, state, and execution traces explicit runtime abstractions. The code is publicly available.