Skip to main content
Glama
Diterex

youtube-research-mcp

by Diterex

get_video_frames

Read-onlyIdempotent

Extract video frames at specific timestamps to see what a YouTube video actually shows, solving the problem that transcripts miss visual details like clicks and dialog values.

Instructions

See what a video actually shows at chosen moments. Returns real images.

Transcripts cannot capture a screen-based tutorial. "Click this, then drag it here" has no referent in text, toolbar clicks are usually silent, typed dialog values are rarely spoken, and auto-captions mangle exactly the technical terms you need. Use this to look at the moments that matter.

The intended workflow is two steps, and doing it in this order is what keeps it cheap:

  1. get_video_transcript(..., include_timestamps=True) to find WHICH moments matter, across as many videos as you like.

  2. get_video_frames(video_id, timestamps=[...]) on just those moments.

The video-only stream is fetched to a temp file first, then every frame comes off it locally. That sounds expensive and is not: video-only means no audio track, so a 31 minute 720p tutorial is about 32 MB and lands in under 10 seconds, and further calls on the same video are instant because the file is kept for the life of the server process (3 videos max, deleted on exit). Asking for many timestamps in ONE call is therefore much cheaper than many calls, and vastly cheaper than one call per frame on different videos.

Args: video_url_or_id: An 11-character video ID or any YouTube video URL. timestamps: The moments to capture, as 'S', 'M:SS' or 'H:MM:SS' strings (e.g. ["4:12", "11:38", "1:02:05"]). Take these from a timestamped transcript. every_seconds: Instead of explicit timestamps, sample evenly this many seconds apart. Use only when surveying an unfamiliar video; explicit timestamps are far cheaper. Ignored if timestamps is given. max_frames: Hard cap on frames returned (1-20, default 6). Every frame costs context, so keep this tight. width: Output width in pixels (320-1920, default 1280). Do not go below about 960 if you need to read menu labels or dialog values. max_height: Source stream height to fetch (default 720, which is enough to read a CAD toolbar and keeps the fetch small). Raise to 1080 only if 720 proves too coarse. quality: JPEG quality, 2 is best and 31 is worst (default 4).

Returns: A list whose first item is a text summary (video title, duration, and the timestamp of each frame in order), followed by one image per timestamp. Frames that could not be captured are reported in the summary text rather than failing the whole call.

Errors: Raises ValueError for bad arguments or unparseable timestamps, and RuntimeError if ffmpeg is missing or the video has no playable stream.

Input Schema

TableJSON Schema
NameRequiredDescriptionDefault
widthNo
qualityNo
max_framesNo
max_heightNo
timestampsNo
every_secondsNo
video_url_or_idYes
Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Beyond the annotations (readOnly, idempotent, openWorld), the description reveals significant behavioral traits: the video-only stream is fetched to a temp file, cached for the server process lifetime (3 videos max, deleted on exit), and multiple timestamps in one call are much cheaper than separate calls. It also discloses failure handling (frames reported in summary) and specific error types.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is long but every sentence earns its place. It is front-loaded with the core purpose, followed by workflow, cost/performance context, detailed parameter semantics, return format, and errors. The structured sections (Args, Returns, Errors) make it easy to scan, and no information is redundant with the annotations or schema.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

For a tool that returns images, the description fully explains the return format (summary text plus images) and how failures are surfaced. It also documents error conditions and performance tradeoffs, covering all the practical context an agent needs to invoke the tool correctly. The presence of an output schema would not add much beyond what is described.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters5/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Although the schema has 0% description coverage, the description's Args section thoroughly explains each parameter: video_url_or_id, timestamps format examples, every_seconds semantics, max_frames cap with cost warning, width guidance for readability, max_height tradeoffs, and quality range. This adds substantial meaning beyond the bare schema.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description opens with a specific verb+resource: 'See what a video actually shows at chosen moments. Returns real images.' It clearly distinguishes the tool from transcript-based alternatives by explicitly stating what transcripts cannot do and positioning this tool for visual verification.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Provides explicit when-to-use guidance: use it to look at moments that matter in screen-based tutorials. It names the sibling alternative get_video_transcript and prescribes a two-step workflow with this tool as the second step, including cost-saving rationale. It also gives a conditional use case for every_seconds and warns against misuse.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Install Server

Other Tools

Latest Blog Posts

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/Diterex/youtube-research-mcp'

If you have feedback or need assistance with the MCP directory API, please join our Discord server