03 · cli-agent · 15 systems
CLI & terminal agentsTerminal-native agents that edit repos, run commands, and commit.
CLI agents run in the shell against a local checkout. They read the repository, plan changes, edit files, execute commands and tests, and often commit — usually with tool protocols like MCP for extension.
Systems in this category
Claude Code
Anthropic · cli-agent
Claude Code is Anthropic's terminal-native coding agent: it plans, edits, runs commands, uses Git and MCP, and sustains long agentic loops.
Codex CLI
OpenAI · cli-agent
Codex is OpenAI's coding agent: an open-source terminal CLI, a desktop app, IDE extensions, and a cloud lane for async tasks, all included in ChatGPT plans.
Aider
Community / OSS · cli-agent
Aider is an open-source, Git-native pair programmer: it maps the repo, proposes commits, and works with any model via BYOK or OpenRouter.
Amp
Amp Inc. · cli-agent
Amp is a terminal-first, multi-model coding agent with subagents, an Oracle second-opinion model, threads, and MCP plugins.
OpenCode
Community / OSS · cli-agent
OpenCode is an MIT-licensed open-source terminal coding agent that is fully BYOK, with optional hosted model access via Zen and Go.
OpenHands
Community / OSS · cli-agent
OpenHands (formerly OpenDevin) is an open-source software engineering agent you can self-host, using LiteLLM or OpenRouter for models.
Goose
Block · cli-agent
Goose is Block's local-first agent runtime with many model providers and extensible tool plugins.
Gemini CLI
Google · cli-agent
Gemini CLI is Google's open-source terminal agent; on June 18, 2026 it stopped serving free and Google AI Pro/Ultra accounts in favor of the Antigravity CLI.
Warp
Warp · cli-agent
Warp is a terminal that ships its own coding agent (Warp Agent / Oz) with plan-bundled credits, so the shell itself runs agentic tasks.
Muse Code
Meta · cli-agent
Muse Code is Meta's terminal coding agent powered by Muse Spark 1.2 — persistent background subagents, a replay-exact event log, workflow orchestration, and a TypeScript SDK over the Muse Session Protocol.
Grok Build
xAI · cli-agent
Grok Build is xAI's terminal coding agent — a full-screen, mouse-interactive TUI (open-source Rust) powered by Grok 4.6, with skills, subagents in Git worktrees, headless mode, and ACP embedding for editors.
Kimi Code
Moonshot AI · cli-agent
Kimi Code is Moonshot AI's single-binary terminal agent — tuned TUI, subagents, conversational MCP setup, lifecycle hooks, and ACP for Zed/JetBrains. Runs Kimi models by default but also speaks Anthropic, OpenAI, and Gemini protocols.
Qwen Code
Alibaba Cloud · cli-agent
Qwen Code is Alibaba's open-source terminal agent — adapted from Gemini CLI and tuned for Qwen3-Coder — with subagents, agent teams, auto-memory and auto-skills, IDE plugins, a desktop app, daemon mode, and SDKs.
Mistral Vibe
Mistral AI · cli-agent
Mistral Vibe is Mistral's open-source CLI coding agent powered by Devstral 2 — terminal-native multi-file automation with IDE extensions for VS Code, JetBrains, and Zed over ACP. Its enterprise counterpart is Mistral Code.
Crush
Charm · cli-agent
Crush is Charm's open-source terminal coding agent — multi-model sessions that can switch providers mid-session, LSP-enhanced context, MCP support, and a polished TUI on every major platform.
Who it's for
Engineers doing hard multi-file work who want the agent alongside their existing editor and git workflow, and teams scripting agents into CI or headless runs.
Tradeoffs
Strong on autonomy and repo scope; weaker on inline editing ergonomics. Pricing splits between subscription bundles (Claude Code) and pass-through or BYOK models (Amp, Aider, OpenCode).
Frequently asked
What is a CLI agent?
CLI agents run in the shell against a local checkout. They read the repository, plan changes, edit files, execute commands and tests, and often commit — usually with tool protocols like MCP for extension.
Who should use a CLI agent?
Engineers doing hard multi-file work who want the agent alongside their existing editor and git workflow, and teams scripting agents into CI or headless runs.
What are the tradeoffs of CLI agents?
Strong on autonomy and repo scope; weaker on inline editing ergonomics. Pricing splits between subscription bundles (Claude Code) and pass-through or BYOK models (Amp, Aider, OpenCode).
Which CLI agents are in the catalog?
15 as of 2026-09-25: Claude Code (Anthropic), Codex CLI (OpenAI), Aider (Community / OSS), Amp (Amp Inc.), OpenCode (Community / OSS), OpenHands (Community / OSS), Goose (Block), Gemini CLI (Google), Warp (Warp), Muse Code (Meta), Grok Build (xAI), Kimi Code (Moonshot AI), Qwen Code (Alibaba Cloud), Mistral Vibe (Mistral AI), Crush (Charm).
What is a terminal coding agent?
A program you run in the shell that acts on your repository — reading code, editing files, running commands and tests, and committing — driven by an LLM. Claude Code, Codex CLI, Grok Build, and Muse Code are the vendor examples; Aider, OpenCode, OpenHands, Goose, Kimi Code, Qwen Code, and Crush are open source.
Can CLI agents run in CI or headless?
Several are designed for it — running as a step in CI or a background job. Check each entry's interfaces field; entries listing api alongside cli generally expose a scriptable surface.
Which terminal agents are open source and BYOK?
Aider, OpenCode, OpenHands, Goose, Kimi Code, Qwen Code, and Crush in the current seed. They bill nothing themselves; you pay the model provider or a router. Grok Build's source is Apache-2.0 but still requires a paid xAI subscription.