Connect an AI assistant
Let Claude, ChatGPT, Cursor or VS Code find your videos, read transcripts and summaries, and comment or share for you. You sign in and allow access; there's no key to copy.
On this page
Vidler has an MCP server, so AI assistants can work with your videos without you pasting anything in: find the video where something was said, read its transcript with timestamps, its AI summary and its comments, see who watched, and (with Read and write access) comment, share, upload and have Vidler AI write documents.
You don't need an API key. The assistant opens Vidler in your browser, you choose the workspace and the access, and you select Allow access.
The server URL is https://vidler.ai/api/v1/mcp.
Connect an assistant
Go to Settings → Integrations, open the Developers tab and find AI assistants (MCP). It shows the Server URL and a tab for each assistant, with one-click links. Or set it up by hand:
Claude Code
Run this in a terminal:
claude mcp add --transport http vidler https://vidler.ai/api/v1/mcp
claude mcp login vidlerOr, inside Claude Code, run /mcp, choose vidler and select Authenticate. Add -s user to the first command to have Vidler in every project.
Claude on the web, Claude Desktop and mobile
- In Claude, open Settings → Connectors and select Add custom connector. The Open Claude settings button in Vidler fills it in for you.
- Name it
Vidler, paste the server URL and select Add. - Select Connect on the new connector, then Allow access in Vidler.
Connectors are shared by Claude on the web, Claude Desktop and the mobile apps, so you connect once. On a Team or Enterprise plan, an Owner adds it first under Organization settings → Connectors.
Cursor
Select Add to Cursor in Vidler, then Install. Or add this to ~/.cursor/mcp.json:
{ "mcpServers": { "vidler": { "url": "https://vidler.ai/api/v1/mcp" } } }Then select Connect next to Vidler in Cursor's MCP settings.
VS Code
Select Add to VS Code (or Add to VS Code Insiders) in Vidler, then Install. Or add this to .vscode/mcp.json in your project:
{ "servers": { "vidler": { "type": "http", "url": "https://vidler.ai/api/v1/mcp" } } }When VS Code asks to authenticate with Vidler, select Allow. It needs VS Code 1.99 or later with GitHub Copilot.
ChatGPT and others
- ChatGPT: on the web, select the plus button, then Add custom MCP server. Name it Vidler, paste the server URL, choose OAuth for Authentication and create it. In business workspaces an admin may need to allow it.
- Other assistants: add the server URL as a remote MCP server. Any assistant that supports MCP sign-in finds Vidler's sign-in by itself.
What an assistant can read
It acts as you, so it reads what you can open in Vidler:
- your own videos and your workspace's;
- videos shared with you from other workspaces, by your email or through Anyone at {your domain} (these are read only for the assistant);
- a video from another workspace whose link you give it, when anyone with the link can watch it.
For each video it can read the transcript with timestamps, the AI summary, chapters, takeaways and action items, and the comments. Ask it things like Find where pricing came up last week or Summarise the onboarding video Bea shared with me: it searches what was said and links to the moment, such as https://vidler.ai/v/aB3dE5fG?t=83.
What the owner hid from viewers stays hidden. If Transcript, Summary or Chapters is off in a video's settings, an assistant that can only watch the video can't read them either.
Allow access
The page that opens shows:
- Who's asking, and where Allow access sends you: a website such as claude.ai, or an app on this computer.
- Recognised by Vidler when it's one of the assistants above. Otherwise an orange warning says Vidler doesn't recognise the app. Continue only if you started connecting yourself. For a website Vidler doesn't recognise, tick I trust … first.
- Workspace: the one it can use. If you belong to several, choose one.
- Access:
- Read only: find and read videos, transcripts, summaries and comments.
- Read and write: also comment, react, share, upload, change titles and folders, and have Vidler AI write documents.
- Admin: offered to owners and admins when the assistant asks for it. It keeps your owner or admin powers, such as other people's private videos.
The assistant acts as you. It sees only the videos you can open, and it can't delete videos, change workspace settings or see your password.
Select Deny access to cancel. Nothing is shared.
See and disconnect assistants
Connected assistants, in the same Developers tab, lists every assistant you've connected. Each row shows its access, when you connected it, when it was last used and where it sends you. Select Disconnect to stop it straight away.
Owners and admins can switch to Everyone to see and disconnect every assistant in the workspace.
An assistant also stops working when:
- the person who connected it leaves the workspace;
- the workspace starts requiring two-factor authentication, and they haven't turned it on;
- the workspace starts requiring single sign-on, and the assistant wasn't connected through it;
- it's used from outside the workspace's allowed networks.
Everything an assistant changes appears in the workspace activity log under API keys and integrations.
If something goes wrong
- "This link points somewhere unexpected": the app asked Vidler to send you somewhere it hadn't registered, so Vidler stopped and shared nothing. Start again from the assistant.
- "This app isn't set up to connect to Vidler": remove Vidler from the assistant and add it again.
- "Needs reconnecting" in Connected assistants: Vidler signed the assistant out for safety, for example because one of its sign-in tokens was used twice. Connect it again from the assistant.
Use an API key instead
Scripts, and assistants that can't sign in this way, can send an API key. The With an API key tab shows the set-up, and API keys, webhooks and developer tools explains keys.