Codewhale
Verifiedby Codewhale
Runs a terminal coding agent with plan and work modes, approval levels, durable goals, agent teams, and MCP, using hosted or local models through Ollama, vLLM, or SGLang.
- Access
- CLI
- Auth
- None
- Skills
- 5
Skills
Plan And Work Modes
Explores read-only in plan mode, edits in work mode, and lets you pick Ask, Auto-Review, or Full Access approval levels.
Headless Exec Runs
Runs a task non-interactively with codewhale exec for scripts and CI, or opens the same runtime in a local web client.
Agent Teams
Coordinates teams of agents with roles kept as readable files in the project, without dumping their instructions into your transcript.
GitHub PR Review
Reviews pull requests from a GitHub Action that runs codewhale review on a pinned release, authorized by a Codewhale account key.
Undo And Restore
Recovers workspace changes with /undo and /restore, and saves sessions with a durable /goal so long work can resume.
Related Agents
MiniMax Code
by MiniMax AI
Runs MiniMax's open-source coding agent in the terminal for developers, with an interactive TUI, headless mcode exec…
Codex CLI
by OpenAI
Runs a coding agent locally in the terminal, editing files and executing commands inside an OS sandbox with…
Factory
by Factory
Runs enterprise coding Droids from the terminal with MCP tools, reusable Missions and Skills, plus a headless exec mode…
Qwen Code
by Alibaba Qwen
Runs an open-source coding agent in the terminal with subagents, agent teams, hooks, skills, plan mode and MCP…