Turn screen recordings and video links into change lists with screenshots an AI can act on.
Maintain this server? Claim it
Call this server's tools from your browser. Nothing to install.
Requests go through mcptry because browsers can't call most MCP servers directly. Header values are sent to the server for that request only and are never stored or logged.
Tool list from our last check. Connect live to refresh it.
Use when the user gives you a video FILE on their computer (a screen recording .mov/.mp4/.webm, a Loom or QuickTime download) that you cannot watch. This server cannot read local files, so this tool returns the exact one-line curl command that uploads the file and prints the change list as Markdown; run it with your shell tool. The result lists every change asked for, with the words said, where on screen, how to check it, a screenshot link of that moment, and the timed transcript.
Path of the video file on the user's computer.
Pick your client. Values in capitals, like YOUR_TOKEN, are yours to fill in.
Run in a terminal
claude mcp add --transport http tellscreen https://tellscreen.com/mcp
Use when the user gives you a video FILE on their computer (a screen recording .mov/.mp4/.webm, a Loom or QuickTime download) that you cannot watch. This server cannot read local files, so this tool returns the exact one-line curl command that uploads the file and prints the change list as Markdown; run it with your shell tool. The result lists every change asked for, with the words said, where on screen, how to check it, a screenshot link of that moment, and the timed transcript.
| Argument | Type | Description |
|---|---|---|
| path* | string | Path of the video file on the user's computer. |
{
"type": "object",
"required": [
"path"
],
"properties": {
"path": {
"type": "string",
"description": "Path of the video file on the user's computer."
}
}
}Use when the user gives you a link to a video (a screen recording, product walkthrough, bug repro or design review: an .mp4/.mov/.webm URL, or a public Google Drive or Dropbox link) and you need to know what it shows or asks for, because you cannot watch video. Tellscreen transcribes it, looks at the screen at the moment each thing was said, and returns Markdown: a summary, every change asked for with the words said, where on screen, how to check it is done and a screenshot link of that moment, then the timed transcript. Waits up to 50 seconds; if it is still working you get a job id for get_change_list. Loom and YouTube pages cannot be fetched: ask the user for the downloaded file instead and use convert_local_video.
| Argument | Type | Description |
|---|---|---|
| url* | string | Direct link to the video file, or a public Google Drive / Dropbox link. |
| title | string | Optional title for the video. |
{
"type": "object",
"required": [
"url"
],
"properties": {
"url": {
"type": "string",
"description": "Direct link to the video file, or a public Google Drive / Dropbox link."
},
"title": {
"type": "string",
"description": "Optional title for the video."
}
}
}Collect the result of record_screen or convert_video_link by job id. Waits up to 50 seconds; call it again while it says it is still working. Also accepts a tellscreen.com/v/ link to fetch the change list of an existing Tellscreen video.
| Argument | Type | Description |
|---|---|---|
| job | string | The job id returned by record_screen or convert_video_link. |
| video_url | string | Or: a tellscreen.com/v/… link (keep any ?agent= in it). |
{
"type": "object",
"properties": {
"job": {
"type": "string",
"description": "The job id returned by record_screen or convert_video_link."
},
"video_url": {
"type": "string",
"description": "Or: a tellscreen.com/v/… link (keep any ?agent= in it)."
}
}
}Use when the user wants to show you something instead of typing it ("let me show you", "watch this", a bug they can reproduce, a design or product review). Returns a recording link for the user to open: they record their screen and talk through what should change, then press Stop. Then call get_change_list with the job id to receive the change list (it waits for them).
No arguments.
{
"type": "object",
"properties": {}
}We read every tool's name, description and inputs the way the model will, looking for instructions hidden from you. A clean result means no warning signs, not a guarantee. Scanned Nov 10, 2026.