Reference
43 AI coding agent CLIs Loopy runs
Loopy isn't a coding agent — it's the agent operating system that orchestrates them. Bring whichever terminal coding agents you already use; Loopy routes each task to the right one and fails over when an agent hits a usage cap.
Pick a default per project in the dashboard, pass --agent to the CLI, or let Loopy choose. Run loopy doctor to see which agents are installed and authenticated on your machine. Agents fall into three groups by how they authenticate.
Uses your existing CLI login (no extra API key)
Already signed in? Loopy drives it as-is.
- Codex
- Claude Code
- Cursor Agent
- Gemini CLI
- GitHub Copilot CLI
- Amazon Q Developer
- Warp Agent
- Antigravity
- Qwen Code
- Kimi Code CLI
- OpenClaw Runtime
- Freebuff
Local login or your own API key
Run on a local session or drop in a provider key.
- Amp (Sourcegraph)
- Cline CLI
- Continue CLI
- Goose
- OpenCode
- OpenHands CLI
- openHarness
- Roo Code
- Kilo Code
- Crush
- Crab Code
- Coro Code
- Command Code
- Codebuff
- Factory
- Factory Droid
- ForgeCode
- Devon
- Every Code
- Neovate Code
- Letta Code
- LogiCoal
- Trae Agent
- Mistral Vibe
- gptme
- Pi
Bring your own API key
Key-based agents you configure once in settings.
- Aider
- Plandex
- SWE-agent
- Backboard R-CLI
- Solo CTO Agent
Why orchestrate instead of picking one?
- Right tool per task — a debug ticket, a docs ticket, and a migration can each run on the agent that's best at it.
- Failover — when an agent nears its usage limit, Loopy hands the turn to a secondary so the build doesn't stall.
- Parallelism — independent slices of a build can run across agents (or named subagents) at the same time.
- No lock-in — your work graph, policies, and evidence live in Loopy; the coding agent is swappable.
Don't see your agent? Loopy registers terminal-native CLIs, so most new agents are a small adapter away. Open an issue on GitHub.
Orchestrate your coding agents
Local-first, on your keys. Assign work; Loopy picks the agent and ships.