Skip to main content
Glama

Squish a video into a timestamped contact sheet

squish_video

Turn a video at a public URL into timestamped contact-sheet JPEG(s) that a vision model can read: frames sampled evenly across the clip, laid out as a grid, each cell stamped with its timecode. Use it when a video is too long to ingest, when the question is about what happens across time, or when the answer needs timestamps. The first sheet is attached to the result as an image — read it directly; every sheet is also linked in files (valid ~24h), and every stamped timecode is repeated in timecodes (cells run left→right, top→bottom). Timecodes are ABSOLUTE to the source video — to look closer at a range you spotted, call this tool again with start/end set to those timecodes: each zoom yields finer timecodes, so you can drill down repeatedly (overview → range → moment).

Tool call template

This tool declares no required parameters. Add optional parameters from the input schema as needed.

{
  "id": 1,
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "arguments": {},
    "name": "squish_video"
  }
}

Input Schema

NameRequiredDescriptionDefault
endNoZoom-window end — same formats as start. Omit to run to the end of the clip; values past the end are clamped.
startNoZoom-window start — seconds (67.5) or a timecode as stamped on a sheet ("1:07", "1:07.3"). Absolute in the source video. Omit to start at 0.
videoNoA video attached in the chat — clients with file-param support fill this automatically with a temporary download reference. Provide either this or video_url.
densityNoGrid density. 3x3 recovers what happened; denser grids (4x4-6x6) recover how it was done. Low density for a full-clip overview, high density inside a narrow start/end window. Default 3x3.
video_urlNoPublic http(s) URL of a video file (anything ffmpeg decodes). Not a YouTube/streaming page — a direct file URL. Provide either this or an attached video, not both.

Output Schema

NameRequiredDescriptionDefault
filesYesSheet URLs, valid ~24h (sheet_ttl_hours).
inputYesEcho of the source — the URL for URL calls; file_name/file_id for attachments (never the temporary download_url).
framesYesFrames sampled across the clip or window.
job_idYes
layoutYesWhere each cell sits on each sheet, in sheet pixels — same order as files, index-aligned with timecodes (cells[n] is the frame stamped timecodes[n]). Published because it cannot be derived: gutters, an outer margin and the branding band mean the grid does not fill the image, so dividing width/height by columns and rows is wrong. Crop by these rectangles to show a single frame. IMPORTANT: these describe the full-resolution sheets at files[], NOT the inline image block — that preview is re-encoded down a size ladder to fit the result budget, so it is often smaller. To crop the inline preview instead, scale every rectangle by (its decoded width / layout[0].width).
sheetsYesNumber of contact sheets produced.
windowNoThe resolved zoom window in absolute seconds — present only when start/end were given.
contractYesAlways "squish-mcp-http-v0".
durationYesSource video duration, seconds.
warningsYes
timecodesYesPer-sheet stamped timecodes, cells left→right then top→bottom — mirrors every label burned into the sheets. Absolute to the source video at every zoom depth.
credits_chargedNoKeyed calls only.
sheet_ttl_hoursYes
credits_remainingNoKeyed calls only.
free_jobs_remaining_todayNoAnonymous calls only.

Schema Changelog

Changes observed during successful MCP inspections.

  1. Changed2 schema fields changed
    • addedOutput schema / properties / layout
      Added value: +{
      +  "description": "Where each cell sits on each sheet, in sheet pixels — same order as files, index-aligned with timecodes (cells[n] is the frame stamped timecodes[n]). Published because it cannot be derived: gutters, an outer margin and the branding band mean the grid does not fill the image, so dividing width/height by columns and rows is wrong. Crop by these rectangles to show a single frame. IMPORTANT: these describe the full-resolution sheets at files[], NOT the inline image block — that preview is re-encoded down a size ladder to fit the result budget, so it is often smaller. To crop the inline preview instead, scale every rectangle by (its decoded width / layout[0].width).",
      +  "items": {
      +    "additionalProperties": false,
      +    "properties": {
      +      "cells": {
      +        "items": {
      +          "additionalProperties": false,
      +          "properties": {
      +            "blank": {
      +              "description": "True when that frame could not be read and the cell was left black. Present on EVERY cell — a missing field means a server older than this contract, not a healthy sheet. Blankness is NOT detectable from the image: every cell is painted black before drawing, so a night shot or a cut to black is pixel-identical to a failed read.",
      +              "type": "boolean"
      +            },
      +            "h": {
      +              "type": "number"
      +            },
      +            "w": {
      +              "type": "number"
      +            },
      +            "x": {
      +              "type": "number"
      +            },
      +            "y": {
      +              "type": "number"
      +            }
      +          },
      +          "required": [
      +            "x",
      +            "y",
      +            "w",
      +            "h",
      +            "blank"
      +          ],
      +          "type": "object"
      +        },
      +        "type": "array"
      +      },
      +      "height": {
      +        "type": "number"
      +      },
      +      "width": {
      +        "type": "number"
      +      }
      +    },
      +    "required": [
      +      "width",
      +      "height",
      +      "cells"
      +    ],
      +    "type": "object"
      +  },
      +  "type": "array"
      +}
    • changedOutput schema / required
      Previous value: -[
      -  "input",
      -  "duration",
      -  "frames",
      -  "sheets",
      -  "files",
      -  "sheet_ttl_hours",
      -  "warnings",
      -  "timecodes",
      -  "job_id",
      -  "contract"
      -]New value: +[
      +  "input",
      +  "duration",
      +  "frames",
      +  "sheets",
      +  "files",
      +  "sheet_ttl_hours",
      +  "warnings",
      +  "timecodes",
      +  "layout",
      +  "job_id",
      +  "contract"
      +]
  2. First observed

TDQS

A4.6/5.0
Behavior4/5

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

Adds real behavioral context beyond the annotations: the first sheet is attached as an image, all sheets appear in `files` with ~24h validity, and timecodes are repeated in `timecodes` with left→right/top→bottom ordering. It also discloses that timecodes are absolute and that zoom calls yield progressively finer stamps. It stops short of stating limits or failure modes (max clip length, unsupported sources beyond 'not a YouTube page'), which keeps it at 4 rather than 5.

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

Conciseness4/5

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

Front-loaded with the core transformation, then usage, then output mechanics, then the drill-down loop—a sensible information order. It is dense and runs long, but nearly every clause carries non-redundant operational detail; only minor trimming is possible.

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?

With an output schema present, annotations covering safety, and full schema coverage, the remaining burden is workflow and return-value interpretation—both covered (attachment vs. `files` links, `timecodes` ordering, TTL, recursive zooming). Nothing an agent needs to invoke and chain this tool correctly is missing.

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

Parameters4/5

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

Schema coverage is 100%, so parameter definitions already carry the format details; the description nonetheless adds semantics the schema cannot—that start/end are ABSOLUTE to the source video and that repeated zoom calls are the intended drill-down mechanism. It also frames density as overview-vs-narrow-window, reinforcing the enum's meaning with a usage heuristic.

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?

States a specific verb and output artifact: 'Turn a video at a public URL into timestamped contact-sheet JPEG(s)', and immediately characterizes the payload (sampled frames, grid layout, per-cell timecode). There is no ambiguity about what comes back or what resource is consumed.

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?

Gives explicit selection conditions ('too long to ingest', 'question is about what happens across time', 'answer needs timestamps') and then the recursive workflow: call again with start/end set to the spotted timecodes, overview → range → moment. This is the strongest form of when-to-use guidance short of naming a sibling.

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

Try in Browser

Glama MCP Gateway

Add one secure layer between your agents and this server.