Wingman-AI
An open-source AI agent platform with local execution, Ollama and LM Studio support, persistent sessions, and a macOS desktop companion. MIT licensed.
Wingman-AI is a self-hosted AI agent platform for work that needs ongoing context and several agents with different roles. It suits developers and teams handling research, support, or recurring operations alongside coding tasks. A lead agent can delegate work to specialized subagents, each with its own workspace and session history.
The runtime runs through a local gateway by default, with a separate local execution mode for isolated tasks. It supports Ollama and LM Studio without requiring provider authentication. Cloud connections include Anthropic, OpenAI, OpenRouter, xAI, and Copilot; choosing those providers sends model requests to their services rather than keeping inference local. The project is open source under the MIT license.
Sessions persist in the gateway across clients and devices, so context can carry between the terminal and browser chat interface. A VS Code extension provides editor access, and the macOS desktop companion can act as a node for running system commands and displaying notifications. Tools can also present interactive components in the web interface.
For repeated work, Wingman-AI supports scheduled routines and webhooks that trigger agents from external systems. Message routing directs incoming requests to a designated agent, while rooms can request parallel responses. Custom agents and reusable skills let teams define domain-specific behavior. MCP tools connect agents to external systems, and hooks support automation before and after tool calls.
