Extend agent-mux to support multiple adapter types beyond subprocess-based execution. This enables integration with sophisticated tools that provide HTTP APIs, WebSocket interfaces, or direct SDK access.

This design proposal covers adding kanban and workspaces views to packages/agent-mux/tui as a thin presentation and command layer over existing packages/kanban and packages/agent-mux/core seams.

This spec defines how amux resolves which LLM provider and model a harness should use, covering both the CLI surface (amux launch, amux run) and the programmatic SDK. It introduces a provider configuration system that abstracts away per-harness differences in how providers and models are specified.

amux launch is a CLI command that launches (or resumes) an interactive or non-interactive coding agent session with full stdin/stdout passthrough. Unlike amux run, which parses and normalizes agent output into the AgentEvent stream, amux launch acts as a transparent proxy — the user interacts directly with the harness as if they invoked it natively.

amux-proxy is a lightweight Python package that bridges between LLM transport protocols. Each instance serves a single purpose: accept requests in one protocol format (the exposed transport) and forward them to a target provider in its native format, translating between the two using LiteLLM(https://github.com/BerriAI/litellm) as the translation engine.