Agents in Solo are AI coding tools that you configure in Settings → Agents and launch inside a project. Regular commands stream terminal output and can also accept terminal input. Terminal-backed agents use terminal input for conversational prompts and follow-up questions.
Agents use your own binaries, your own auth, your own skills, your own configuration. Solo is, at it's core, a very good terminal. We do not interfere with the agents installed on your computer. If a new CLI agent comes out tomorrow, you can simply add it in Solo without waiting for us to update anything!
Agents vs. regular commands
| Commands | Agents | |
|---|---|---|
| Purpose | Run dev servers, build scripts, workers | AI-assisted coding and development |
| Interaction | Stream output and accept terminal input | Type prompts, have a conversation |
| Sidebar section | Commands | Agents |
| Status indicators | Starting / Running / Stopping / Stopped / Exited / Failed | Working / Planning / Idle while running; Starting / Stopping / Stopped / Exited / Failed otherwise |
| Closing | Stop button (no confirmation) | Remove agent context-menu action (running agents always confirm; stopped-agent confirmation can be suppressed) |
| Summaries | Excluded from automatic summaries | Optional, when available |
Supported agent tools
Solo ships with built-in support for these AI coding agents:
- Claude
- Codex
- Amp
- Gemini
- OpenCode
- Copilot
- Kimi
- Antigravity
You can also add custom agent tools — any CLI-based AI tool that works in a terminal.
How agents appear in the sidebar
Agents show up in the Agents section for a project. Section order is user-configurable, so it isn't always literally at the top.
Each agent instance:
- Starts with the tool name. A custom display name or an enabled terminal-title update can change what the sidebar shows.
- Shows a state indicator. While running, Working and Planning animate and Idle is static; non-running terminal-backed agents use lifecycle indicators.
- Can display an auto-generated summary line below its name when a summary is available.
You can have multiple agent instances running simultaneously — one per task or topic.
Nested agents
If one agent starts another agent through Solo's MCP tools, the child can appear nested under the parent. Child rows can be collapsed under the parent, and collapsed parents can show how many subagents are hidden.
Getting started
- Configure which agent tools are available in Settings → Agents. See Setting up agent tools.
- Launch an agent from a project. The New agent/terminal shortcut defaults to
Cmd+Ton macOS andCtrl+Ton other platforms, and can be changed or disabled in Settings → Hotkeys. See Launching agents for all launch paths. - For a complete first run, follow Build, review, and land one change. When a task has several independent lanes, use the agent orchestration workflow.