usewhale/Whale

Whale — blazingly fast, terminal-first AI coding agent for DeepSeek. ~98% prompt cache hit rate, 1M context, MCP tools, dynamic workflows.

Whale – AI‑powered coding assistant for DeepSeek

What it is – Whale is a terminal‑first, DeepSeek‑optimized AI agent that can read, edit, and run code, browse the web, and invoke tools. It ships as a single binary (or npm package) and offers a TUI, a one‑shot CLI, and a headless mode for CI/CD.

Key capabilities

  • Prompt caching (~98 % hit rate) – Re‑uses previously sent context so most interactions cost only a few cents.
  • DeepSeek‑native – Tailored to DeepSeek’s 1 M‑token context, tool‑calling, and pricing model.
  • Dynamic workflows – Write JavaScript scripts that orchestrate multiple agents (parallel research, synthesis, etc.). Compatible with Claude Code scripts.
  • MCP, Skills & Plugins – Connect to thousands of MCP servers (file ops, git, web, databases), load community “skills” (code review, git worktree), or add custom plugins and hooks.
  • Multiple interfaces – Interactive TUI (whale), quick CLI (whale ask "…"), and headless mode for automation.

Installation

# npm (cross‑platform)
npm install -g @usewhale/whale
# macOS via Homebrew
brew install usewhale/tap/whale
# Linux one‑liner
curl -fsSL https://raw.githubusercontent.com/usewhale/Whale/main/scripts/install.sh | sh
# Windows PowerShell
irm https://raw.githubusercontent.com/usewhale/Whale/main/scripts/install.ps1 | iex

After installing, run whale setup to store your DeepSeek API key, then start the agent with whale.

Typical workflow

  1. Launch the TUI (whale).
  2. Type a natural‑language request (e.g., “refactor this function”).
  3. Whale reads the relevant files, calls the DeepSeek model, may invoke tools (shell, git, web search), and returns edited code.
  4. For scripted pipelines, place a .whale/workflows/*.js file and enable Dynamic Workflows in the config.

Project status – The repository is no longer maintained. The README directs users to the official DeepSeek Harness for a maintained alternative. The code remains usable, but there will be no further updates or official support.

Who it’s for – Developers who want a fast, low‑cost AI coding assistant that works directly from the terminal and can be embedded in CI pipelines, without the overhead of a full IDE integration.

Related

  • Project
    codewhale-hq/Codewhale

    Codewhale is an open-source terminal-based coding agent that automates repository tasks, file editing, and command execution using a wide range of AI models.

  • Project
    wavetermdev/waveterm

    An open-source, AI-integrated terminal that combines a command-line interface with context-aware AI assistance, durable SSH sessions, and built-in graphical editors.

  • Project
    openkursar/hello-halo

    Halo AI Workstation is an open‑source desktop app that combines a chat‑style AI workbench with autonomous “Digital Humans”—AI agents that can run on schedules, keep memory, and collaborate in distributed teams. It supports multiple LLM back‑ends (Claude, ChatGPT, etc.), offers built‑in document editing, web browsing, terminal control, and integrates with IM platforms (Feishu, WeChat, WeCom) for remote hand‑offs. The software runs locally on macOS, Windows, Linux, and has mobile remote clients, requiring no backend service.

  • Project
    iAmCorey/Wake

    Wake is a native Rust‑based desktop app that aggregates, indexes, and lets you search the local conversation histories of many AI coding agents (Claude Code, Codex, Cursor, etc.). It offers a unified UI, full‑text search, one‑click resume of sessions, remote syncing over SSH, and a read‑only MCP/CLI for agents to query the history.

  • Project
    WrongStack/WrongStack

    A standalone, open-source AI coding agent with durable codebase memory, multi-agent orchestration, and a built-in toolbox for autonomous software development.