MCP App Store
by KeenEthics
Castify (Screencastify) icon

Castify (Screencastify) MCP Server for ChatGPT

By Castify (Screencastify)

Overview

Castify (Screencastify) is a leading video creation and sharing platform used by 10M+ people and 2,000+ organizations. This connector gives ChatGPT access to your Castify video library, transforming your content into action-oriented AI workflows. Ask in plain language about the videos you've created or collected (any video can be added), and ChatGPT can pull views, comments, and analytics to show what videos are working and what isn't. Then act on it right away: build custom ordered playlists of related videos or organize your library into folders by subject, unit, or whatever structure fits how you teach. Because it runs inside ChatGPT, you can combine Castify with your other connected tools, like turning a playlist into a ready-to-send Gmail draft.

How to connect

  1. ChatGPT: Open the Castify (Screencastify) listing in the ChatGPT app directory and choose Connect. ChatGPT listing Castify (Screencastify) exposes 23 tools here.

Tools

add_youtube_videos_to_library

ChatGPT
Adds existing YouTube videos to the user's Screencastify library by link. Nothing is uploaded to or published on YouTube: the library entry points to the video on YouTube, and the video stays there. Once added, a video can be shared, commented on and tracked like any other video in the library. Use it when the user wants to bring YouTube videos into their library. It does not apply to video files (a different tool handles those) or to Loom videos. Input and limits: - Each video is a YouTube watch, youtu.be or embed link, or the 11-character video ID. Playlists, channels, Shorts and live streams are not supported. - A title is required for each video. It names the video in the library and can be changed later. - At most 10 videos per call. Each one succeeds or fails on its own, and the result lists both. Side effects: - Creates one library entry per video. This counts against the organization's video quota, and the call fails if the quota is reached. Calling the tool again adds the videos again. - Videos are private to the user unless sharing is given. Sharing takes the same entries as add in update_video_sharing, follows the same rules, and gives the recipients access to every video in the call. - With folderId, the videos go into that folder if the user can add to it. Otherwise they stay in the library root and the result says so (inFolder: false). Requires a signed-in user. Organization API keys cannot add videos.

create_folder

ChatGPT
Create Folder Creates a new folder to help organize videos. Optionally nests it inside an existing folder and/or moves a set of videos into it immediately. CRITICAL CONTEXT RULES 1. User Context Only: Folders are always owned by the authenticated user, so this tool can only be used when operating on behalf of a specific user. Organization API keys (which have no specific user) cannot create folders. 2. Parent Folder Ownership: If parentFolderId is provided, it must be a folder owned by the authenticated user; this is enforced automatically. 3. Video Ownership: Every video listed in videoIds must be owned by the authenticated user; this is also enforced automatically. Use execute_search with ownerId set to your own user ID first if you need to confirm which videos you own.

create_video_uploads

ChatGPT
Create Video Uploads Starts the upload of one or more video files from the user's computer. This tool does not send any file: it creates a draft video for each file and returns an uploadUrl that you send the file to yourself. To add a YouTube video use add_youtube_videos_to_library instead. CRITICAL CONTEXT RULES 1. User Context Only: This tool can only be used when operating on behalf of a specific user. Organization API keys (which have no specific user) cannot create videos. 2. You must be able to read the file and make an HTTP request (for example with a shell). If you can't, tell the user you can't upload files from here. Do not call this tool if you can't send the file. 3. Measure the size. sizeBytes must be the exact size of the file, measured from the file itself (for example stat). The upload is rejected if the file is any other size, and the upload URL can't be reused for a different file. 4. Send the file: one PUT of the whole file to uploadUrl, with the Content-Type that was declared and no other credentials, for example curl -X PUT -H "Content-Type: <contentType>" --upload-file <file> "<uploadUrl>". A 200 or 201 means the file was received. To resume an interrupted upload, ask how far it got with PUT to the same URL with Content-Length: 0 and Content-Range: bytes */<total>; a 308 response carries a Range header with the bytes received. Continue from the next byte with Content-Range: bytes <start>-<end>/<total>. 5. The `uploadUrl` is a secret. Use it only for that file. Never show it to the user or write it anywhere. 6. Deadline: send the file before expiresAt. After that, start again. 7. Processing: after the upload the file is checked and moved to the user's storage, and the video shows up in their library when that is done. This can take a few minutes. Tell the user it is being processed; don't promise it is ready. Files that are not valid videos are rejected. Use get_video_upload_status with the videoIds to check on them. 8. Quota and limits: the organization's video quota applies. Each file must be an allowed type and at most 5 GiB. A user can only have a limited number of uploads in progress at once; if the call fails because of that, wait for the current ones to finish. 9. Sharing: videos are private to the user unless sharing is given. It takes the same entries as add in update_video_sharing. Do not share unless the user asked you to. 10. Partial results: each file succeeds or fails on its own. Calling the tool again starts new uploads, so only retry the files that failed.

export_video

ChatGPT
Export Video Starts exporting a video as a downloadable file. Exporting takes a while, so this only starts it: the result tells you how far along it is and when to check again. Use get_video_export with the returned exportId until the status is "ready", then give the user the downloadUrl. To get a video's transcript use get_video_transcript instead. CRITICAL CONTEXT RULES 1. User Context Only: This tool can only be used when operating on behalf of a specific user. Organization API keys (which have no specific user) cannot export videos. 2. Permissions: You can only export videos you own or have been given collaborator or contributor access to; this is enforced automatically and the call fails otherwise. 3. Asking again is safe: If an identical export is already running or was finished recently, you get that export back (alreadyStarted is true) instead of starting another one. 4. Download links are temporary. Do not store or reuse them; call this tool again for a fresh one. 5. Failures: If the export fails, the call reports why. Calling this tool again starts a new attempt.

export_video_document

ChatGPT
Exports a generated document of a Screencastify video as a Word (.docx) or PDF file and returns a temporary download link for it. The file is the one the web app downloads: the original document, with its screenshots. Use it when the user wants a generated document (a how-to guide, a bug report, a study guide or a custom write-up) as a file to download or share. Input and limits: - format is "docx" or "pdf". - Omit documentId to export the video's newest document, or pass it to export a specific one. Finding the newest document needs the user to be the video's owner or a collaborator. A documentId works for anyone who may see that document. - Only the original document is exported, not translated versions. Documents are created with generate_video_document. Side effects: - Builds the file and keeps a temporary copy in storage. The link works for a few minutes, and anyone who has the link can download the file until it expires. The copy is deleted within a day. - Each call creates a new copy and a new link. Earlier links keep working until they expire. Requires a signed-in user. Organization API keys cannot export documents.

generate_video_captions

ChatGPT
Starts generating captions for a Screencastify video, which also gives the video a transcript. Generation runs in the background and takes a few minutes, so a call only starts it and returns a taskId. Once it has finished, get_video_transcript returns the captions. Use it when the user wants captions or a transcript for a video that does not have them yet. A video that already has captions is refused. Input and limits: - Only Screencastify videos are supported, not YouTube videos. The language is detected from the audio. - The user must be the video's owner or a collaborator. Contributors and viewers are refused. Side effects: - Queues a captions job for the video and, when it finishes, adds the captions to the video. - Counts against the organization's captions limit once the job finishes. The call is refused when the plan does not include captions or the limit is already used up. - Nobody is emailed when the captions are ready. - Asking again while captions are being generated for the video returns that running task (alreadyStarted is true) and does not start another. Requires a signed-in user. Organization API keys cannot generate captions.

generate_video_document

ChatGPT
Starts generating a written document from a Screencastify video's captions, such as a how-to guide, a bug report, a study guide or a custom write-up. Generation runs in the background and takes a few minutes, so a call only starts it and returns a taskId. The finished document is saved with the video. Use it when the user wants a written document based on what is said in a video. Input and limits: - Only Screencastify videos are supported, not YouTube videos. A video can have up to 5 documents, and a sixth is refused. - The user must be the video's owner or a collaborator. Contributors and viewers are refused. - contentType is "how-to-tutorial", "bug-ticket", "study-guide" or "custom". A custom document needs instructions. Side effects: - Queues a document job for the video and, when it finishes, saves the document with the video. It is not published to the video's watch page, and nobody is emailed when it is ready. - Documents are written from English captions. If the video has none, a captions job is started first and the document follows once it finishes. Those captions count against the organization's captions limit when they finish. - Counts against the organization's document limit once the document finishes. The call is refused when that limit is already used up. - Asking again while a document of the same contentType is being generated for the video returns that running task (alreadyStarted is true), even if the other options differ. Asking again for a different contentType, or after the document has finished, generates another document. Requires a signed-in user. Organization API keys cannot generate documents.

get_assignment_analytics

ChatGPT
Get Assignment Analytics Returns a single Submit assignment's analytics: how many of the requested submissions have been submitted (completion), how they scored against the assignment's maximum points, and how many times they've been viewed and by how many different people. A measure with no data behind it yet — for example an assignment nobody has submitted to or viewed — comes back as null rather than 0 or omitted. CONTEXT RULES 1. User Context only: This tool requires a User Context. Organization API keys are not supported. 2. Access control: You can only see analytics for an assignment you own — this tool checks that itself before fetching analytics. An assignment you can't access is reported the same way as one that doesn't exist. 3. Assignments, not videos: This is for Submit assignments (topics in search results). For a regular video, use get_video_analytics. 4. Freshness: Numbers can lag a fresh submission, score or view by a couple of minutes. 5. New assignments: An assignment that exists but hasn't been picked up by analytics yet returns all measures as null — this is expected, not an error.

get_auth_context

ChatGPT
Get Auth Context Fetch the current authenticated context (user identity, organizations, roles, and accessible resources). CRITICAL INSTRUCTIONS - Always call this tool first if you need to look up an organization ID by name, as many other tools (like search) require an organization ID. - Organization Token Note: When using an organization token, there will be no "user" metadata present in the response.

get_video_analytics

ChatGPT
Get Video Analytics Returns a single video's engagement analytics: how much of it viewers typically watch, how they scored on its quiz (if it has one), and how many times it's been viewed (total and unique views). A measure with no data behind it yet — for example a video with no quiz, or one nobody has watched — comes back as null rather than 0 or omitted. CONTEXT RULES 1. User Context only: This tool requires a User Context. Organization API keys are not supported. 2. Access control: You can only see analytics for a video you own — this tool checks that itself before fetching analytics. A video you can't access is reported the same way as one that doesn't exist. 3. New videos: A brand-new video with no analytics activity yet simply returns all fields as null — this is expected, not an error.

get_video_export

ChatGPT
Get Video Export Checks on an export started with export_video. While its status is "processing", wait the suggested retryAfterSeconds and call this tool again. Once it is "ready", give the user the downloadUrl. CRITICAL CONTEXT RULES 1. User Context Only: This tool can only be used when operating on behalf of a specific user. Organization API keys (which have no specific user) cannot check exports. 2. Permissions: You can only check exports of videos you own or have been given collaborator or contributor access to; this is enforced automatically and the call fails otherwise. 3. Failures: If the export failed, the call reports why. Call export_video again to start a new attempt. 4. Download links are temporary. Do not store or reuse them.

get_video_transcript

ChatGPT
Get Video Transcript Returns a video's transcript as a WebVTT (VTT) file. Small transcripts come back inline, attached as a file named after the video's title (also in text), so you can read, summarize or quote them directly. Larger ones come back as a short-lived download url instead; give that link to the user, as you cannot read the file yourself. CRITICAL CONTEXT RULES 1. User Context Only: This tool can only be used when operating on behalf of a specific user. Organization API keys (which have no specific user) cannot read transcripts. 2. Permissions: You can only get transcripts for videos you own or have been given collaborator or contributor access to; this is enforced automatically and the call fails otherwise. 3. Language: Omit language to get the default transcript. If a language is requested and the video has no transcript in it, the call fails rather than falling back to another language. 4. Download links expire a few minutes after they are issued. Do not store or reuse them; call this tool again for a fresh one.

get_video_upload_status

ChatGPT
Get Video Upload Status Checks where uploads stand: still waiting for the file, being processed, ready, held for the user to fix something, or rejected. Use it after sending the files of create_video_uploads, after starting an import with import_loom_videos, and when the user asks how an upload is going. CRITICAL CONTEXT RULES 1. User Context Only: This tool can only be used when operating on behalf of a specific user. Organization API keys (which have no specific user) cannot check uploads. 2. Own uploads only: Only uploads the user started can be checked. Any other video is reported as not found. 3. Processing takes a while. After the file is sent the video is checked and moved to the user's storage, which can take a few minutes. An import that the backend carries out (a Loom video) can also report how far it has come (progressPercent). Don't poll in a tight loop: check once or twice, then tell the user it is still being processed and offer to check again later. 4. Relay `message`: each entry has a plain-language message. Blocked uploads (for example a full or disconnected Drive) are kept for a while and retried once the user fixes the problem; tell them what to fix. Rejected uploads ended without a video; the user needs to start the upload again. 5. `uploadUrl` is a secret: it is only returned while the upload is waiting for its file, so an interrupted upload can be resumed. Never show it to the user. 6. Ready videos: offer the watchUrl.

import_loom_videos

ChatGPT
Copies Loom videos into the user's Screencastify library, given their Loom share links. Use it when the user wants to keep a Loom video in Screencastify. It does not apply to video files (a different tool handles those) or to YouTube videos (add_youtube_videos_to_library). How it works: for each link, Screencastify retrieves the video from Loom the same way a viewer with that link would, and respects Loom's own settings. Only public Loom videos can be copied: a video that is private, password protected, deleted or has downloads turned off is rejected, and the reason is reported by get_video_upload_status. Nothing is fetched from any site other than loom.com. Input and limits: - Each link must be a Loom share or embed link, such as https://www.loom.com/share/<id>. Any other link is refused, and so is the whole call, so nothing is imported. - At most 100 links per call. Links to the same video are combined into one video. Side effects: - Starts one import per video and creates a library entry for it. Imports run in the background and take from seconds to a few minutes, so a video is not ready when the call returns. get_video_upload_status reports progress, and the watch link once it is ready. - The videos are placed in the user's "Loom Imports" folder, which is created if it does not exist. They are first titled "Loom video" and receive Loom's title and captions when the import finishes. - Counts against the organization's video quota, and a user can only have a limited number of uploads and imports in progress at once. Calling the tool again starts new imports. - Videos are private to the user unless sharing is given. Sharing takes the same entries as add in update_video_sharing, follows the same rules, and gives the recipients access to every video in the call. Requires a signed-in user. Organization API keys cannot import videos.

list_assignments_by_metric

ChatGPT
List Assignments By Metric Ranks the Submit assignments you own by a single measure — completion rate, average score, total views, or unique viewers — from highest to lowest or lowest to highest. Good for "which of my assignments have the lowest completion" or "which have the most unique viewers". CONTEXT RULES 1. User Context only: This tool requires a User Context. Organization API keys are not supported. 2. Scope: Only assignments you own within orgId are eligible to appear — this tool resolves that set itself, you don't need to pass assignment IDs. 3. No data yet: Assignments are never left out for lacking data. Never-viewed assignments have 0 views; assignments with nothing to average yet have a null completion rate or average score and are listed after those with a value. 4. Assignments, not videos: For regular videos, use list_videos_by_metric.

list_shareable_groups

ChatGPT
List Shareable Groups Lists the groups the user can share videos with (for example classes or teams), across all the organizations they belong to. Use it to find a group's ID before calling update_video_sharing. CRITICAL CONTEXT RULES 1. User Context Only: This tool can only be used when operating on behalf of a specific user. Organization API keys (which have no specific user) have no groups to share with. 2. Who sees which groups: Admins see every group in their organization; members see the groups they belong to; restricted users (often students) see their own groups only if the organization allows student-to-student sharing. 3. Names, not IDs: When talking to the user, always use the group's name (add the organization name if two groups share a name). The ID is only for tool calls.

list_submissions_by_score

ChatGPT
List Submissions By Score Ranks the submissions on a specific Submit assignment by score, from highest to lowest or lowest to highest, with each submitter's name and email. Good for "who scored lowest on this assignment" or "who are the top performers". CONTEXT RULES 1. User Context only: This tool requires a User Context. Organization API keys are not supported. 2. Scope: You can only rank submissions on an assignment you own within orgId. An assignment you can't access is reported the same way as one that doesn't exist. 3. Scored submissions only: Scores are compared as a fraction of the assignment's maximum points, so submissions without a score — and every submission on an assignment with no maximum points — are excluded. An empty result doesn't mean there are no submissions. 4. Submitter identity: submitterUserId and submitterEmail are null when the submitter only entered a name.

list_users_by_score

ChatGPT
List Users By Score Ranks the people who took a specific video's quiz by their score, from highest to lowest or lowest to highest. CONTEXT RULES 1. User Context only: This tool requires a User Context. Organization API keys are not supported. 2. Scope: You can only rank quiz results for a video you own within orgId. A video you can't access is reported the same way as one that doesn't exist. 3. Identified responders only: An anonymous quiz response can't be attributed to a person, so it's excluded entirely — it never appears as a null user ID.

list_videos_by_metric

ChatGPT
List Videos By Metric Ranks the videos you have access to by a single engagement measure — average completion, average quiz score, total views, or unique views — from highest to lowest or lowest to highest. CONTEXT RULES 1. User Context only: This tool requires a User Context. Organization API keys are not supported. 2. Scope: Only videos you own within orgId are eligible to appear — this tool resolves that set itself, you don't need to pass video IDs. 3. Folder scoping: Pass folderId to rank only a specific folder's direct contents. Omit it to rank across everything you own in orgId.

update_folder

ChatGPT
Update Folder Renames a folder, moves it under a different parent, and/or moves videos into it. Only the fields you provide are changed; omitted fields are left as-is. CRITICAL CONTEXT RULES 1. User Context Only: This tool can only be used when operating on behalf of a specific user. Organization API keys (which have no specific user) cannot manage folders. 2. Folder Ownership: You can only update folders you own; this is enforced automatically and the call fails if the folder doesn't belong to you. To avoid a failed call, use execute_search (filtering by ownerId, set to your own user ID) to confirm ownership beforehand. 3. Parent Folder Ownership: If parentFolderId is provided, it must also be a folder owned by the authenticated user; this is enforced automatically. 4. Video Ownership: Every video listed in videoIds must be owned by the authenticated user; this is also enforced automatically. Each listed video will be relocated into this folder; videos already in the folder that aren't listed are left untouched.

update_video

ChatGPT
Update Video Renames a video and/or edits its description. Only the fields you provide are changed; omitted fields are left as-is. To change who a video is shared with use update_video_sharing instead. CRITICAL CONTEXT RULES 1. User Context Only: This tool can only be used when operating on behalf of a specific user. Organization API keys (which have no specific user) cannot update videos. 2. Permissions: You can only update videos you own or have been given collaborator or contributor access to; this is enforced automatically and the call fails otherwise. 3. Title changes update the video's metadata only; the underlying file is not renamed.

update_video_sharing

ChatGPT
Update Video Sharing Changes who a video is shared with by adding and/or removing sharing entries: people (by email), groups (by name), the whole organization (optionally just one type of seat), an email domain, or link access. To change a video's title or description use update_video instead. CRITICAL CONTEXT RULES 1. User Context Only: This tool can only be used when operating on behalf of a specific user. Organization API keys (which have no specific user) cannot change sharing. 2. Permissions: Only the video's owner or a collaborator can change sharing; contributors cannot. This is enforced automatically and the call fails otherwise. 3. Groups: Get group IDs from list_shareable_groups and pass the ID in the call, but always tell the user the group's name. Only groups the user can share with are accepted. Organization entries share with people in the video's own organization (the school or company account that owns it); you never say which organization. appliesTo narrows it to one kind of seat. Sharing with everyone in the organization at a level above view is allowed. 4. Access levels: view < copy < edit < manage. Use the lowest level that does what the user asked; people who only need to watch get view. 5. Incremental: remove entries are applied first, then add entries. You never need to send the full sharing list. The owner cannot be added or removed. Sharing rules (e.g. restrictions on who a school user may share with) are enforced by the server and the call fails if a change is not allowed. 6. Notifications: Share notifications are off unless notify is true. Only set it when the user asked for recipients to be notified. 7. Watch link: The result includes watchUrl, the link people use to open the video. Offer it to the user so they can send it to recipients (it only works for people who now have access).

Capabilities

Writes

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/castify-screencastify"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/castify-screencastify)