🤙 yeaboi.ai

Best friend to engineers and agents — plans, standups, retros, performance & reporting for your team, plus cost, digests and security posture for the AI agents working alongside it. All from your terminal.

Two worlds, one command: scrum for your team, and cost, digests & security posture for your agents.

🚀 Quick Start

Recommended: uv or pipx

The most reliable way to install — pulls the full dependency tree from PyPI and isolates it in its own environment:

uv tool install yeaboi          # or: pipx install yeaboi
yeaboi --setup                  # configure your API key
yeaboi                          # launch the interactive TUI

Note on names: the package was previously published as scrum-agent. It is now yeaboi on PyPI, matching the command. A final scrum-agent release remains as a thin redirect that installs yeaboi, and the legacy scrum-agent command still works as an alias for that release — but new installs should use yeaboi.

Voice input needs no separate install. Double-tap Space in any text field and yeaboi offers to set dictation up in place — it installs the packages into the environment it is already running in, downloads the speech model with a progress bar, and drops straight into recording. No restart, no second terminal. (yeaboi --install-voice does the same thing headlessly, for CI and dev containers.)

It is kept out of the base install rather than bundled because the speech engine (ctranslate2, onnxruntime) publishes wheels only for 64-bit macOS, glibc ≥ 2.28 Linux and Windows — making it a hard dependency would break pip install yeaboi outright on Alpine/musl, older glibc, 32-bit and armv7 hosts. On those platforms yeaboi says so instead of offering an install that cannot work.

Optional extras can still be requested at install time if you prefer:

uv tool install "yeaboi[voice]"                # 🎤 dictation, pre-installed rather than on demand
uv tool install "yeaboi[all-providers]"        # OpenAI, Google, and Bedrock providers
pipx install "yeaboi[voice]"                   # equivalent with pipx

Voice input transcribes on-device with faster-whisper — no API key, works with every LLM provider (Anthropic, Bedrock, …). On macOS/Windows it is fully self-contained (the sounddevice wheel bundles PortAudio). On Linux, also install the system library: sudo apt install libportaudio2 — yeaboi will not run sudo for you. A small Whisper model downloads on first use (~140 MB for the default base; set VOICE_MODEL to tiny/small/medium/large-v3 to trade size for accuracy).

Homebrew is not supported. A required dependency (sqlite-vec) ships no source distribution, which Homebrew's source-build model can't handle, so brew install yeaboi is intentionally disabled. Use uv tool install or pipx install above instead.

From source

git clone https://github.com/dinho149/yeaboi.ai.git
cd yeaboi.ai
make install        # installs uv, creates venv, installs dependencies
make env            # creates .env from .env.example — add your API key
make run            # launch the CLI

Headless / CI mode

yeaboi --non-interactive --description "Build a todo app" --output json
yeaboi --non-interactive --description @project-brief.txt --output html --team-size 5

✨ Features

🖥️ Full-screen TUI — Animated splash, mode selection, pipeline progress, dark/light themes 🧠 Smart Intake — Extracts answers from your project description, asks only what's missing — or feed it a whole quarterly roadmap with Roadmap Intake 🔄 Seven modes, one command — Planning, Daily Standup, Retro, Planning Poker, Performance (beta), Reporting, Team Analysis 🤖 Agents too, not just humans (beta) — a robotic-duck landing split opens the Agents family: what your AI coding agents cost, a daily digest of what they did, and a security audit of your agent setup — computed locally from Claude Code session logs, transcripts never leave your machine 🔌 37 tools — GitHub, Azure DevOps, Jira, Confluence, Notion, local codebase scanning, and more 📤 5 export formats — Markdown, HTML, JSON, Jira sync, Azure DevOps Boards sync 🤖 5 LLM providers — Claude (default), GPT, Gemini, AWS Bedrock, or fully local & keyless with Ollama 🧩 Every surface — TUI, CLI subcommands, MCP server, and a Claude Code plugin, with feature parity enforced in CI 💾 Session persistence — SQLite-backed sessions plus a saved-runs hub for past standups, retros, and reports 🛡️ Guardrails — Input/output validation, human-in-the-loop review at every stage

📖 Full Documentation

Getting started, the full CLI reference, every mode in depth, integrations, architecture, and deployment guides: yeaboi.ai/docs

📄 License

MIT License. See LICENSE for details.