Linchpin

Linchpin is an open-source runtime that lets you deploy and manage AI agents on your own infrastructure. It functions as a self-hosted platform where you can run AI models from various providers—either through cloud services like OpenRouter or locally via Ollama. Each agent operates in its own Docker container, which comes pre-equipped with practical tools like bash, file reading/writing, and web search capabilities. This setup allows the AI to perform tasks within a controlled environment without being able to escape its container.

The system is built around a modular architecture. An orchestrator manages session loops, processing event logs and handling model interactions. You can extend an agent's capabilities by connecting Model Context Protocol servers or custom HTTP tools. Security is handled through encrypted vaults that manage credentials, ensuring secrets are never stored in plain text. The entire platform is designed for transparency and control, running on your hardware with your API keys.

Linchpin is particularly useful for developers and teams who need to build, test, and run AI agents in a private, customizable environment. It's ideal for scenarios like developing agent-based applications, experimenting with different AI models, or deploying production agents where data privacy and infrastructure control are important. The codebase is Apache-2.0 licensed and designed to be simple enough to understand and modify for specific needs.