eat-pray-ai/yutu
The AI-powered toolkit that grows your YouTube channel on autopilot.
yutu – AI‑powered YouTube automation toolkit
What it is – yutu is a Go‑based command‑line tool that also ships an MCP (Model‑Control‑Protocol) server and an AI‑agent mode. It talks to the YouTube Data, Analytics and Reporting APIs and lets a language model drive the whole YouTube workflow: uploading videos, editing titles / descriptions, managing comments, playlists, channel branding, live streams, etc.
Key components
- CLI – a rich set of sub‑commands (
video upload,comment list,playlist create, …) that map directly to YouTube API actions. - Agent mode – launches an LLM‑backed “Miffy” agent (default Gemini‑3.8‑flash, but any OpenAI‑compatible model works). The agent receives a prompt, calls the CLI internally, and can reason about SEO, growth strategies, and channel branding.
- MCP server – exposes the same functionality over a standard MCP interface (stdio or HTTP) so other AI‑assistant platforms (Claude, Codex, VS Code, Cursor, etc.) can invoke YouTube operations as “skills”.
- Skills package – a reusable skill definition (
skills/youtube) that describes the YouTube domain for any AI agent that consumes OpenCode skills.
How you get it – pre‑built binaries (GitHub releases, Homebrew, WinGet, npm, Docker) or go install. Installation scripts verify cryptographic attestations.
Setup steps
- Create a GCP project and enable the YouTube Data API (required) plus optional Analytics/Reporting APIs.
- Create OAuth Client ID (Desktop app) and download
client_secret.json. - Run
yutu auth --credential client_secret.jsonto obtainyoutube.token.json. - Set environment variables (
YUTU_CREDENTIAL,YUTU_CACHE_TOKEN, etc.) or use the--credential/--cacheTokenflags.
Typical usage
# upload a video with automatic title/description generation via Gemini
yutu agent --provider google --model gemini-3.8-flash --api-key $GEMINI_KEY
# direct CLI upload
yutu video upload --file myvideo.mp4 --title "My Title" --description "..."
# start MCP server for other tools
yutu mcp --mode http # listens on 8216 by default
Why it matters – Managing a YouTube channel is tedious and API‑heavy. By wrapping the YouTube APIs in a CLI and exposing them to LLMs, yutu lets developers and content creators automate repetitive tasks, apply AI‑driven SEO recommendations, and integrate YouTube actions into broader AI agents or workflows.
Documentation – README, FEATURES.md, INSTRUCTION.md, and auto‑generated Go docs (pkg.go.dev). Contribution guidelines are provided.
Related
- Project
- Project
- Project
- Project
- Project