solo-agent/solo

What it solves

Solo provides a unified, local-first workspace for coordinating multiple AI coding agents. It eliminates the fragmentation caused by running agents in separate terminal tabs or chat transcripts, preventing the need to re-explain context for every new session and turning untracked conversations into structured, reviewable tasks.

How it works

Solo operates through three local layers: a Next.js frontend, a Go-based API server for persistence and communication, and a daemon that manages agent subprocesses. It integrates with existing agent CLIs (such as Claude Code, Codex, and OpenCode) by communicating via stdin/stdout, providing these agents with shared memory, prompts, and collaboration tools.

Who it’s for

Developers who use multiple AI coding agents and need a centralized environment to manage their work, track tasks, and maintain long-term memory across different agent sessions.

Highlights

• Channel-based Coordination: Shared rooms for humans and agents to chat, share files, and manage channel-scoped teams.
• Threaded Task Boards: Kanban-style boards where every task has an attached discussion thread, subtasks, and artifacts.
• Thinking Mode: Ability to branch conversations into focused lines of reasoning without losing the parent discussion context.
• Agent Observability: Dashboards to track live agent runs, inspect session transcripts, and monitor usage trends.
• Agent Team Templates: Pre-defined workflows or custom-described roles to quickly set up agent teams.
• Long-term Memory: Agent-specific MEMORY.md files that maintain context across sessions.