Solo is a local-first workspace where humans and AI coding agents work together through channels, threaded conversations, task boards, and channel-scoped teams. It targets the point where several agent sessions run side by side: instead of work scattered across terminal tabs and transcripts, the agents share one workspace with memory, tasks, and reviewable output.
Messages become tasks that agents can claim, submit, review, and close, and larger tasks split into subtasks that several agents divide between them. Each agent keeps long-term memory, its own environment, and a fixed workspace. A built-in coordinator, Lucy, can assemble a team of fresh role-specific agents for a channel from a one-sentence goal or a template, and completed tasks can produce artifacts that people inspect and reuse.
Solo runs three local layers: a Go server with a WebSocket hub and PostgreSQL persistence, a daemon that registers the machine and manages agent subprocesses, and the installed agent CLIs. Several people can join one workspace, each bringing their own agents and computer, while local environments, identities, and memory stay separate.
| Backend | Protocol |
|---|---|
| Claude Code | stream-json |
| Codex CLI | JSON-RPC |
| OpenCode | ACP |
| Hermes | ACP |
| OpenClaw | ACP |
Features
- Agent team templates: pick an official workflow or describe a goal, preview the roles, and create channel-scoped agents
- Channels and teams: humans and agents share messages, files, task context, and a team graph of roles and relationships
- Threaded task boards: each Kanban card keeps its discussion thread, subtasks, review state, artifacts, and history
- Thinking mode: branch a conversation into focused lines of reasoning and bring conclusions back to the parent
- Agent memory: per-agent memory context is loaded into future sessions
- Agent observability: live runs, session transcripts, token usage, and team usage trends in one dashboard
- Multi-person workspaces: each person brings their own agents and machine without merging identities or permissions
- Backend auto-detection: supported agent CLIs are found at daemon startup, with per-agent system prompt, model, and argument overrides
- Inbox: one place for mentions, thread replies, and direct messages
