Your AI Coding Agents Need a Desk, Not More Tabs

What it does Solo is a local-first workspace that turns scattered terminal sessions with AI coding agents into a persistent team environment. A Go server and local daemon manage agent subprocesses over stdin/stdout, while a Next.js frontend provides humans and agents with shared channels, Kanban tasks, and file attachments. It is essentially ambitious glue: it wraps the agent binaries already on your machine and injects collaboration tools into their otherwise isolated runs.

The interesting bit Instead of treating AI agents as one-shot commands, Solo frames them as long-lived coworkers with channel-scoped roles, an inbox, and a MEMORY.md file that survives across sessions. The system auto-detects supported backends—Claude Code, Codex, OpenCode, Hermes, OpenClaw—from your PATH and speaks their respective protocols so they can share a workspace without knowing about each other. It bills itself as a workspace, not a company simulator: agents are remembered and reviewed, but the goal is finished artifacts, not org-chart theater.

Key highlights

• Runs locally on your own machine: Go API server, daemon, Next.js UI, and PostgreSQL persistence.
• Agents get persistent memory, direct mentions, and channel-scoped team graphs showing roles and relationships.
• Tasks follow a Kanban flow (todo to done) with threaded discussions, subtasks, review stages, and reusable artifacts.
• “Thinking mode” branches a channel conversation into focused reasoning lines without losing the parent context.
• Auto-detects agent CLI binaries and adapts to stream-json, JSON-RPC, or ACP protocols at daemon startup.

Verdict Worth a look if you are already juggling multiple local AI agent CLIs and want one place to coordinate them. Skip it if you do not have supported agents installed or you are looking for a managed cloud service; Solo demands a local Go, Node, and Docker stack.