Provides tools to retrieve Bilibili video metadata (title, author, description, parts, CIDs) and timestamped transcripts/subtitles for individual parts.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@bilibili-video-reader-mcp Get the part list and Chinese subtitles for this Bilibili video: https://b23.tv/BV12345"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
A read-only ChatGPT/Codex plugin built as Skill + stateless MCP server, now targeting Cloudflare Workers so it can run without Render or a credit card.
get_bilibili_video_parts(url) — title, author, description, all P numbers and CIDs.
get_bilibili_part_transcript(url, part, language, max_chars) — official/AI subtitle tracks and timestamped transcript for one P.
The Worker calls Bilibili's metadata/player/subtitle endpoints directly. No video download is needed for subtitle mode.
Related MCP server: Bilibili MCP Server
Create/sign in to a Cloudflare account.
Open Workers & Pages → Create → Import a repository (wording may vary slightly).
Connect GitHub and select limuzi23/bilibili-video-reader-mcp.
Keep the repository root as the root directory.
Cloudflare should detect wrangler.jsonc automatically.
Use the default deploy command: npx wrangler deploy.
Deploy.
The Worker name must remain bilibili-video-reader-mcp because it matches wrangler.jsonc.
After deployment, Cloudflare will give you a URL similar to:
https://bilibili-video-reader-mcp.<your-subdomain>.workers.devHealth check:
https://bilibili-video-reader-mcp.<your-subdomain>.workers.dev/healthMCP endpoint:
https://bilibili-video-reader-mcp.<your-subdomain>.workers.dev/mcpFor public subtitles, no runtime secret is required.
If a subtitle is only visible while logged in, the Worker can read an optional Cloudflare secret named BILI_COOKIE. Configure it only inside Cloudflare's encrypted Variables and Secrets UI. Never paste raw cookies into ChatGPT and never commit them to GitHub.
ChatGPT → Settings → Security and login → enable Developer mode.
Open Plugins, press +, and register the deployed URL ending in /mcp.
Copy the technical ID from the created connection. It begins with plugin_asdk_app_.
Run python scripts/configure_plugin.py plugin_asdk_app_YOUR_ID if the plugin packaging flow requires an app reference.
The OpenAI plugin manifest lives at .codex-plugin/plugin.json and the workflow Skill is in skills/bilibili-video-reader/.
npm install
npm run devDeploy from a local terminal if desired:
npm run deployEndpoints:
GET /health
MCP /mcpThe current production path is:
ChatGPT Skill
↓
Cloudflare Workers /mcp
↓
Bilibili metadata/player/subtitle APIs
↓
structured multi-P metadata + timestamped subtitlesThe old server/ Python implementation remains in the repository only as a reference/legacy implementation; Cloudflare does not use it.
v2.1 focuses on multi-P metadata + official/AI subtitles. It does not yet download audio/video, run Whisper ASR, or OCR source code from video frames. Those are planned only after this subtitle path is verified end-to-end.
This server cannot be deployed
YouTube transcripts, search, channel browsing, and playlists for AI agents via MCP.
Bilibili public video, article, dynamic, creator, comments, reactions, and transcript tools.
Track Bilibili creators and get the latest updates on videos, dynamics, and articles. Fetch user p…
Fetch transcripts, subtitles, chapters, metadata and frames from YouTube and 10+ video platforms