Channel Brains

Channel Brains is a local, stdio MCP server that indexes publicly available YouTube captions into a searchable SQLite FTS5 database. It runs on your computer, needs no API key, and preserves timestamped links back to the original videos.

It is deliberately small and local:

• No hosted service, Docker, browser dashboard, embeddings, or LLM runtime
• No YouTube API key, database server, or cloud account
• One local SQLite database per user, protected by a cross-process ingestion lock
• One channel ingestion job at a time, with resumable per-video progress
• Search results cite the original video and timestamp

Install

Requires uv.

After the public v0.1.0 release, run it directly from GitHub:

uvx --from "git+https://github.com/Pu11en/[email protected]" channel-brains-mcp

The release is published from Pu11en/channel-brains.

For development:

uv sync --extra dev
uv run channel-brains-mcp

The server communicates only through standard input and output. Do not run it as an HTTP service.

First workflow

1. Add the server to an MCP client using one of the configurations below.
2. Call create_brain with a supported YouTube channel URL, such as https://www.youtube.com/@OpenAI.
3. Poll get_brain_status until the brain is ready, paused, or failed.
4. Use search_brain to retrieve timestamped caption matches.

The initial ingestion scans the complete channel listing so it can select up to 50 videos by view count. Caption availability and YouTube rate limits determine how much can be indexed.

MCP tools

Channel Brains exposes exactly these six tools:

Tool	Purpose
create_brain	Validate a channel URL, persist a brain, and start local ingestion.
get_brain_status	Read progress, counts, selection method, and errors without network access.
list_brain_videos	Page through indexed, skipped, pending, and failed video records.
search_brain	Search local FTS5 caption chunks and return timestamped YouTube citations.
get_video_transcript	Page through stored caption chunks for one indexed video.
delete_brain	Remove a completed brain and all its local records. Active ingestion is refused.

Local data and privacy

By default, Channel Brains stores its SQLite database and lock file in the platform’s user-data directory:

• Windows: %LOCALAPPDATA%\channel-brains
• macOS: ~/Library/Application Support/channel-brains
• Linux: ~/.local/share/channel-brains

Set CHANNEL_BRAINS_HOME before launching the MCP server to use another location:

CHANNEL_BRAINS_HOME=/path/to/channel-brains-data uvx --from "git+https://github.com/Pu11en/[email protected]" channel-brains-mcp

Captions and search indexes stay on the local machine. The only network requests are public YouTube requests made by yt-dlp and caption URL retrieval during ingestion.

MCP client configuration

All examples use the published Pu11en/channel-brains release command.

Hermes Agent

Add this entry under mcp_servers in your Hermes configuration, then restart Hermes:

mcp_servers:
  channel_brains:
    command: uvx
    args:
      - --from
      - git+https://github.com/Pu11en/[email protected]
      - channel-brains-mcp
    timeout: 120
    connect_timeout: 60

Hermes discovers the six tools at startup and registers them with an mcp_channel_brains_ prefix.

Claude Code

Register it at user scope:

claude mcp add -s user channel-brains -- \
  uvx --from "git+https://github.com/Pu11en/[email protected]" channel-brains-mcp

Confirm it is available:

claude mcp list

Codex CLI

Add this to ~/.codex/config.toml:

[mcp_servers.channel-brains]
command = "uvx"
args = ["--from", "git+https://github.com/Pu11en/[email protected]", "channel-brains-mcp"]

Restart Codex after saving the file.

OpenCode

Add a local MCP server entry to your OpenCode configuration:

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "channel-brains": {
      "type": "local",
      "command": [
        "uvx",
        "--from",
        "git+https://github.com/Pu11en/[email protected]",
        "channel-brains-mcp"
      ],
      "enabled": true
    }
  }
}

Restart OpenCode to load the server.

Development and verification

uv sync --extra dev --locked
uv run ruff check .
uv run pytest
uv build

The test suite is offline except for the deliberately separate manual live YouTube smoke test.

Limitations

• Public YouTube captions can be unavailable, expired, restricted, or rate-limited.
• Search is lexical SQLite FTS5 search, not semantic search or an answer-generation system.
• The server indexes captions only. It does not download videos, reuse video footage, or create a knowledge graph.
• A channel can contain many videos. The first local ingestion may take time.

License

MIT. See LICENSE.