HipoCap

HipoCap is an open-source security and governance layer for AI agents and large language model (LLM) applications. It enables teams to securely manage and monitor tool and API calls made by LLMs, providing real-time control over what agents can execute and how they interact with external systems. The primary purpose of HipoCap is to reduce operational and security risks when deploying AI agents into production environments.

The platform offers fine-grained, real-time role-based access control (RBAC) for tools, APIs, and actions invoked by LLMs, allowing permissions to be defined, updated, and enforced without redeploying models. HipoCap includes prompt injection protection by inspecting tool calls and contextual inputs, helping prevent unauthorized or unintended behavior driven by malicious or manipulated prompts. It also delivers full observability into agent activity, with detailed logs, traces, and execution metadata so teams can audit, debug, and optimize agent workflows. Because it is open source, HipoCap can be self-hosted and integrated into existing infrastructure and CI/CD pipelines.