Open Source · Apache License 2.0

Grow YouTube faster
without busywork.

Plan, publish, optimize, and respond from one repeatable workflow,
so every upload gets sharper metadata, stronger packaging, better
session paths, and timely community follow-up.

$ npm i -g @eat-pray-ai/yutu
View on GitHub
Autopilot
Channel ops
Boost
Views & subs
AI-Native
Agent / MCP / Skill
78 Tools
All resources
zsh

Features

Every channel workflow, scriptable and agent-ready.

Stage unlisted uploads, sharpen titles and thumbnails, sync playlists and captions, moderate comments and live chat, or hand the entire growth loop to an AI agent.

Unlisted staging to public release pipeline

Stage, optimize, and publish

Upload to an unlisted queue, enrich SEO metadata, attach thumbnails and captions, and route into playlists in one flow.

Scheduled publishing cadence and release checklist

Keep every release on schedule

Coordinate scheduled uploads, multilingual captions, and series playlists from repeatable CLI scripts or CI jobs.

Thumbnail, title, and metadata refresh comparison

Refresh packaging for higher CTR

Update titles, first-two-line descriptions, tags, custom thumbnails, and channel branding without opening YouTube Studio.

Channel catalog audit and metadata uplift

Audit and upgrade your back catalog

Inspect video stats, search rankings, captions, and playlist structure to spot and fix underperforming uploads.

Pinned comment thread and fast creator reply

Engage viewers while attention is fresh

List recent comment threads, post release-hour prompts, and reply to top questions directly from your terminal or agent.

Comment and live chat moderation controls

Keep comments and live chat clean

Flag spam, set comment moderation status, manage live broadcasts, and enforce live chat bans before noise takes over.

Miffy AI agent orchestrating YouTube MCP tools

Let Miffy or your MCP client run the loop

Launch the built-in Miffy agent or connect any MCP client to 78 tools across all 29 YouTube resource types.

Read, Write, and Destructive tool safety tiers

Delegate with explicit safety tiers

Separate read, write, and destructive operations so agents verify resource IDs first and ask before deleting anything.

Quick Start

Start with one growth loop

1

Install

$ npm i -g @eat-pray-ai/yutu
$ yutu version

Put yutu where planning, publishing, and optimization already happen: terminal, scripts, CI, or an agent workspace.

2

Authenticate

Create OAuth credentials in Google Cloud.

$ yutu auth -c client_secret.json
# Your browser has been opened to an authorization URL.

Authorize once to run automated workflows across your channel.

3

Run it

$ yutu channel list --for mine
$ yutu mcp
$ yutu agent --api-key YOUR_KEY

Run the next upload, packaging review, comment pass, or assistant-driven channel audit.

YouTube Skill

Give agents your growth playbook

Extend any AI agent with domain knowledge, multi-step workflows, and SEO best practices across all 29 YouTube resource types.

View SKILL.md
Safety guardrails

Separates read, write, and destructive tools—verifies targets and confirms before deletions.

Common workflows

Chains video uploads, thumbnails, captions, playlists, comments, and live broadcast steps.

Growth & SEO tips

Applies proven guidance for titles, descriptions, tags, thumbnails, and audience engagement.

Ready to try yutu?

Build a repeatable YouTube operating system

Ask an AI agent to set up yutu

›Install and configure yutu by following https://raw.githubusercontent.com/eat-pray-ai/yutu/main/internal/tools/skillgen/setup.md

Or set up yutu manually.

Install the yutu executable

Choose the package manager or runtime that matches your environment.

Node.js / npm

Install the CLI globally when Node is already part of your tooling.

$npm i -g @eat-pray-ai/yutu

Docker

Run yutu in a container with OAuth files mounted.

$docker run --rm -it -u $(id -u):$(id -g) -v $(pwd):/app -p 8216:8216 ghcr.io/eat-pray-ai/yutu:latest

Go install

Install the latest binary directly from the Go module.

$go install github.com/eat-pray-ai/yutu@latest

Linux script

Use the release installer script for Linux environments.

$curl -sSfL https://raw.githubusercontent.com/eat-pray-ai/yutu/main/scripts/install.sh | bash

macOS / Homebrew

Use Homebrew on macOS, or fall back to the same script on Linux.

$brew install yutu

Windows / WinGet

Install yutu on Windows from the WinGet package.

$winget install yutu

Integrate yutu with agents

Connect yutu to any MCP-compatible client.

$yutu mcp
$npx skills add eat-pray-ai/yutu --skill youtube