MindsHub

MindsHub is an open-source platform designed to host, manage, and run AI agents in production environments. It provides a complete infrastructure layer that abstracts away the complexity of agent orchestration, allowing developers to focus on building and deploying autonomous agents without dealing with underlying system details. The platform supports a range of open-source agents, including OpenClaw, NanoClaw, Anton, and Hermes, each offering different capabilities for task execution, reasoning, and tool use. One of the core features is the Model Router, which intelligently selects the appropriate large language model (LLM) for each step of an agent’s workflow based on cost, latency, and capability requirements. This ensures that simple queries use cheaper, faster models while complex reasoning tasks invoke more powerful ones. The router can be configured to switch models dynamically, optimizing both performance and expenses. Agents operate within Scratchpads, which are protected, reproducible execution environments that isolate code and data. Each agent’s execution is sandboxed, preventing unintended side effects and ensuring repeatability. This is critical for debugging, auditing, and running untrusted or third-party agents safely. To support long-running interactions, MindsHub provides a Memories system that enables persistent, cross-session recall. Agents can store facts, user preferences, and prior outcomes, allowing them to maintain context across multiple turns or even separate conversations. This memory is stored in a structured format and can be queried or updated by the agent during execution. Security is handled through the Credentials Vault, an encrypted secrets store. Agents never have direct access to raw credentials such as API keys, database passwords,