Skip to main content
Glama

Clip a video or stream

queue_video

Submit a YouTube, Twitch, Kick, or local video to find highlight moments, crop them to vertical, add captions, and return a job ID for tracking.

Instructions

Hand Clips Kitty a YouTube, Twitch or Kick link and it finds the moments worth posting, crops them to vertical and captions them, all on this computer. Returns a job id; processing a long stream takes a while.

Tool call template

Replace each placeholder with a value matching the input schema. Only required parameters are included; add optional parameters as needed.

{
  "id": 1,
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "arguments": {
      "url": "<url>"
    },
    "name": "queue_video"
  }
}

Input Schema

NameRequiredDescriptionDefault
urlYesYouTube, Twitch or Kick link
focusNoWhat the person wants the clips to be about or to include, in their own words: a topic they talk about, a moment that happened, a time range ("1:35-1:55", "near the end"), or a style (funny moments, laughing, hype, reactions). Pass their sentence as written, keeping words like "make sure" and "at least two": those make it a must-have. It adds weight to those moments and never removes others, so "avoid X" is not something it can do.
forceNoProcess again even if this video was done before
sportNoThe video is a match or a game: its moments (soccer: goals, saves, cards; basketball: dunks, threes, blocks, game winners, and the crowd, bench and courtside reactions) are found from the crowd, the commentary and the scoreboard, one clip per moment with its build-up and reaction, framed to follow the play. Not with gaming or podcast.
gamingNoA game stream: the streamer's webcam in one half and the game in the other, or the game filling the screen when there is no webcam. The streamer is whoever speaks in sync with the audio, never the biggest face. Not with vertical_live, podcast or longform.
podcastNoMulti-camera podcast footage: framing cuts per shot
captionsNoBurn captions into the clips. On unless set to false.
hashtagsNoHashtags every clip from this video must carry, in its title line and description. Use when they name a tag they want on all of them; do not set metadata clip by clip afterwards instead.
longformNoMake horizontal 1920x1080 video instead of vertical clips. short_clips = up to 60s. clips_140 = up to 140s, for X. highlights = one best-of video, 8-20 min. edited_stream = the whole stream with the downtime removed. Leave unset for normal vertical clips.
pipelineNoAn installed, enabled SDK pipeline. For a Gaming detector also set gaming=true. Use its actual id, never guess.
max_clipsNoCap clips from this video
min_scoreNoQuality bar, 0-100
watermarkNoName of a saved branding profile to put on every clip. Names are matched loosely; a wrong one lists the real ones.
long_clipsNo61-180s vertical clips instead of 10-60s. Still 9:16 for TikTok and Shorts. For a wide 16:9 video use longform.
caption_styleNoHow the burned-in captions look. Set only what was asked for; anything left out keeps the saved setting.
gaming_layoutNoThe existing Gaming layout and webcam/game boxes, validated by the engine. Omit to use Gaming's automatic layout detector. Set only from the user's chosen layout.
vertical_liveNoThe video is a livestream that was already vertical (9:16) when streamed: keep its own layout, no face tracking or reframing. Refused if the video is not 9:16. Not with podcast or longform.
publish_when_doneNoPlatforms to publish every clip to as soon as processing finishes, e.g. ["youtube"]. Use this when they ask you to process a video AND publish it: the clips do not exist yet while you are replying, so there is nothing to publish until the job completes.

Schema Changelog

Changes observed during successful MCP inspections.

  1. First observedv0.1.0

TDQS

A3.8/5.0
Behavior4/5

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

With no annotations supplied, the description carries the full burden and does add real context: processing happens on this machine, the call is asynchronous (returns a job id), and long streams take a while. It stops short of stating prerequisites (engine running), failure behavior, or that clips do not exist until the job completes—that detail is buried in the publish_when_done parameter instead.

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?

Two sentences, front-loaded with what the tool does and the input, then the async return contract. Every clause earns its place and nothing is repeated from the schema.

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

Completeness4/5

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

For an 18-parameter, nested-schema, no-output-schema tool, the description covers purpose, input domain, and the return contract (job id) well. It omits the mutual-exclusion relationships between sport/gaming/podcast/vertical_live and does not point at the polling sibling, leaving small but real gaps at this complexity level.

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

Parameters3/5

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

Schema description coverage is 100% and the parameter docs are unusually rich (focus, sport, highlights, longform, publish_when_done all explain their own semantics). The description adds nothing about parameters beyond the URL domain, so baseline 3 is correct.

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

Purpose4/5

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

Names a concrete verb chain and resource: takes a YouTube/Twitch/Kick link, finds postable moments, crops to vertical, captions, runs locally. The input domain (URL only) implicitly distinguishes it from queue_local_file, but no sibling is named explicitly, so differentiation is inferred rather than stated.

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

Usage Guidelines3/5

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

"Returns a job id; processing a long stream takes a while" implies the async pattern and thus queue_status/job_status, but the agent must infer that. There is no explicit when-to-use vs queue_local_file, clip_captions, or the pipeline path, and no exclusions stated in the description itself.

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