Model-agnostic by design
Claude CLI/API, Z.ai GLM, Ollama, LM Studio, vLLM, and any OpenAI-compatible endpoint keep the same memory, MCP tools, and operating surface.
Run OpenAgent with Claude, GLM, Ollama, LM Studio, vLLM, or any OpenAI-compatible provider, then install the Agent Server, CLI Client, and Desktop App as separate downloads.

OpenAgent is one system, but you install each surface independently. Most setups run the Agent Server on the host where the agent should live, then add either the CLI Client or the Desktop App from a separate download.
The persistent runtime in openagent/. This is the actual agent process with models, MCP tools, memory, channels, scheduler, and install or auto-update behavior.
A separate terminal client for connecting to any running OpenAgent Gateway. Use it when you want fast operational access without the desktop UI.
The Electron control surface for chat, memory, MCP configuration, model setup, and graph exploration. It connects to a running Agent Server and is downloaded independently.
Install openagent-framework on the machine that should keep memory, channels, models, and schedules alive.
Use the CLI Client for terminal-first operations, the Desktop App for visual control, or both.
Swap Claude, GLM, Ollama, LM Studio, vLLM, or any OpenAI-compatible endpoint while keeping the same tools, memory, and channels.