M
MCP Directory CURATED DIRECTORY
911 MCP servers indexed
中文
← Back to directory
V

VideoNote MCP

Community
Video links → multi-format notes
How to install ↓ Compare GitHub source repository ↗
Category
Other #124 of 230
Stars
★ 102 Popular
Transport
stdio (local process)
Runtime
Python 3.11+
Credentials
Optional API key
License
MIT
Last commit
Tools
6
53FMRS · D

VideoNote-MCP is an open-source project (MIT licensed) maintained by HuangYincan that packages the end-to-end "video link → multi-format notes" pipeline as an MCP Server plus a Claude Code Skill; it is not an official product of any upstream platform. By default the Agent in the current conversation writes the notes, with a configured LLM serving only as a fallback when the Agent cannot view images; tools can be decoupled and used on demand, with no backend service required. It suits Claude Code users who need note workflows combining transcription, visual understanding, and danmaku/comments.

Strongest · Documentation 14/20 Weakest · Reliability 8/20

Reliability
8/20
Security and permissions
9/20
Maintenance
11/20
Documentation
14/20
Setup experience
11/20
Why each score
Reliability 8/20
The static review can only read the README and the doc index; the server manifest is not cached, and the repository as supplied contains no verifiable CI workflow or committed tests covering the MCP init and tool-list handshake. The pipeline depends on external downloaders, FFmpeg, Whisper models, and third-party platforms (Bilibili/Douyin/YouTube), where any single failure breaks the end-to-end flow. Positively, the README declares health_check for FFmpeg/DB/whisper readiness and cleanup(task_id, dry_run=True) as inspect-before-delete, which is reasonable diagnosability, but these are prose claims only, with no execution evidence; error handling, rate limits, and anti-bot edge cases are undocumented. Per the static calibration (reliability capped at 12 absent verifiable execution evidence) and applying deductions for unproven paths, 8.
Security and permissions 9/20
No real tokens or secrets appear in the README, and there is no sign of malware, covert exfiltration, or fabricated maintainer identity, so no red-line drop to 0-4 is triggered. However, the security boundary is under-disclosed: platform QR-code login (Bilibili, Douyin) implies use of user account credentials without stated storage location, scope, or revocation; the optional fallback LLM key's storage and destination are unexplained; downloading videos and scraping danmaku/comments is outbound network activity. cleanup defaults to a global factory-reset path; dry_run exists but no consistent confirmation gate for destructive operations is stated. Installation pulls uvx videonote@latest from PyPI, an uncontrolled-version supply-chain risk. Missing disclosure is a deduction, hence 9.
Maintenance 11/20
MIT license, a CHANGELOG, CONTRIBUTING, and a dev->main branch flow with PR review are documented, and the README claims a required green CI smoke on dev, yet no actual CI config, release tags, or issue-response history is visible in the supplied material. With 61 stars and 0 open issues and listing on Glama, some visibility exists, but the data is a single snapshot; commit cadence and a security-response channel (no SECURITY.md or vuln-report path seen) cannot be judged. Maintenance intent is visible while governance and versioning evidence is thin, so 11.
Documentation 14/20
Documentation is well layered: the README is an overview, with a doc index, an architecture doc, a user manual (four install methods, setup wizard plus CLI, environment variables, updates, security), and a CHANGELOG, plus two end-to-end real-world cases with full process records, artifact screenshots, and output directory structures; a tool table covers list_tasks/cleanup/health_check and the task layout. Gaps: full parameter signatures, return shapes, failure semantics, and rate/quota limits for the MCP tools are not enumerated in the visible material; platform login, backend model download size/time, cost, and privacy boundaries are unspecified; the actual security and update section contents are unreachable (only linked headings are visible). Usable docs with hidden assumptions and troubleshooting gaps: 14.
Setup experience 11/20
The README gives a three-step main path: claude plugin marketplace add plus plugin install, then /videonote-setup, with uvx videonote@latest requiring no backend service and a default path that needs no LLM key because the in-conversation agent writes the note. Deductions: prerequisites (Python 3.11+, FFmpeg, Whisper model download) are not consolidated into a preflight checklist in the README; uvx latest introduces version drift; platform QR login and the optional LLM key are extra manual steps; no client configuration example (e.g. an mcpServers JSON snippet) is shown. Because the static material contains no verifiable execution evidence (no CI, no committed tests), setup is capped at 15; net score 11.

Static review · not runListed 2026-09-10

Read the FMRS scoring method →

Fit and risk

What it can accessReads local filesWrites / deletes local filesUses the networkConnects to a database

Best for

  • Individual users who want structured, multi-format notes from a single video link quickly
  • Users working in Claude Code who are willing to have the current conversation Agent write notes directly
  • Note workflows that combine speech transcription, visual understanding, danmaku, and comments from multiple sources
  • Users who need task management, multi-video parallelism, and batch output

Not for

  • Users whose clients do not support MCP / Claude Code Skill (the source only describes a Claude Code Skill and MCP Server form)
  • Users who do not want to install local dependencies such as FFmpeg and whisper or download videos
  • Scenarios needing a hosted backend service rather than local execution (this project requires no backend service to start)

Required permissions

  • Read video links and download the corresponding media locally into task directories (note_results/{task_id}/raw/)
  • Write generated artifacts locally (transcripts, notes, screenshots, frames, subtitle exports, etc.) into note_results/{task_id}/gen/
  • Read and write the local SQLite database video_note.db (video_tasks global task index)
  • Invoke local components such as FFmpeg and whisper
  • Optional platform QR-code login (Bilibili, Douyin, etc.) and optional fallback LLM API key configuration

Risks and side effects

  • Downloads video, cover, and other media files, consuming local disk space (cleanup keeps configuration and models by default)
  • whisper transcription runs locally, so speed and resource use depend on machine performance
  • Optionally enabling the fallback LLM requires providing an API key, creating key-management risk
  • Platform QR-code login stores session state, involving account-credential risk
  • Cleanup operations may delete task data; check first with cleanup(task_id, dry_run=True)

Setup

Before you start

Runtime：Python 3.11+

1) Install via the plugin marketplace: claude plugin marketplace add HuangYincan/VideoNote-MCP, then claude plugin install videonote@videonote (the MCP launch command uvx videonote@latest automatically pulls the latest version from PyPI). 2) During installation Claude Code prompts for defaults item by item (style/transcription engine/video understanding/comments, etc.); after installing, run /videonote-setup in the session to finish configuration. 3) (Optional) Configure the fallback LLM key, platform QR-code login (Bilibili, Douyin, etc.), or run the CLI wizard with ! videonote setup; the default path does not require an LLM. 4) Restart the session and tell the agent "make notes for this video" with a link.

Check that it works

After install, run /videonote-setup in the session, restart, then call health_check; if FFmpeg, database and whisper all report ready, the MCP server is connected.

Troubleshooting

  1. Run health_check to verify FFmpeg / database / whisper readiness
  2. Confirm FFmpeg is installed locally, or media processing and transcription will fail
  3. If transcription fails, check that whisper is available and the model has been downloaded
  4. If tasks are missing or titles mismatch, use list_tasks to enumerate the global index (including semantic titles)
  5. Before cleaning up, run cleanup(task_id, dry_run=True) to confirm the impact
  6. If installation does not take effect, restart the session and run /videonote-setup to complete configuration

Things to try

Once connected, you can ask your AI assistant things like:

  • Make a note from this video: https://www.bilibili.com/video/BV1r8nMz4EAj/
  • Use prepare_note_material on this link, then refine the note based on the subtitles
  • Merge my segmented meeting recording and run speaker diarization
  • List all tasks, then clean up the test task

Tools 6

prepare_note_material writes
Produce material from one link (download, transcription, visual understanding, danmaku/comments) for the current Agent to write notes from.
generate_note writes
Fallback generation path: uses the configured LLM when the Agent cannot view images.
process_media writes
Media processing tool, e.g. action="merge" to merge segmented recordings, action="diarize" for speaker separation.
list_tasks read-only
List all tasks (global index, with semantic titles).
cleanup destructive
Clean up per task (pass task_id) or globally (factory reset, no argument).
health_check read-only
Check FFmpeg / database / whisper readiness.

Use cases

Study and exam prep: end-to-end plus video understanding plus subtitle-based refinement to fully cover a course
Meeting minutes: merge segmented recordings → diarize for speaker separation → meeting_minutes style
Lecture close reading: generate end-to-end, then have the agent refine from the full transcript and fill in details by chapter
Video appreciation: enable danmaku plus comment integration so notes include an "audience views" section
Media processing only: use process_media alone for merging, speaker separation, etc.

Supported clients

Claude Code

Listed from the project's documentation, not tested by this site.

Overview

VideoNote-MCP packages the entire "video link → multi-format notes" pipeline as an MCP Server plus a Claude Code Skill: give the agent a link and it automatically runs download → speech transcription → visual understanding → danmaku/comments, with notes written by the Agent in the current conversation by default (a configured LLM is only a fallback when the Agent cannot view images). It can be used end-to-end (one link → one note) or decoupled, taking generation, material, task, and media-processing tools as needed; no backend service needs to be started. It supports mainstream platforms (e.g. Bilibili, Douyin, YouTube) and can export Markdown, subtitles (srt / vtt / json), and LaTeX mathnote PDF among other formats.

Similar servers

BioMCP 75 · B

One binary. One grammar. Evidence from the biomedical sources you already trust.

★ 659 · Tools 11 Compare with this →
Compare with other MCP servers

Source revision fb9376bc91ba Data synced 2026-10-19 Read the FMRS scoring method