Trint MCP Server

Connect MCP-compatible AI assistants to your Trint account to search files, read transcripts and play media.

The Trint MCP (Model Context Protocol) server lets MCP-compatible AI assistants — such as Claude Desktop, the Claude web app, and ChatGPT — work directly with the content in your Trint account. Once connected, the assistant can search your files, read transcripts, browse workspaces, resolve quote timecodes, and play your media inline.

All requests run against your own Trint user account, authenticated as you, and respect the permissions already set on your files and workspaces. The server is read-only apart from a single in-app AI Live suggestions action.

This is not the REST API

The MCP server is a separate integration surface. It authenticates with OAuth 2.0 as a signed-in user, so it does not use the API keys that the rest of the Trint API relies on.

Connecting

The server speaks the MCP Streamable HTTP transport and authenticates with OAuth 2.0. It advertises authorization-server discovery per RFC 9728, so most MCP clients can complete the sign-in flow automatically once you add the server URL.

Use the URL that matches your tenant — see US vs EU Tenant if you are not sure which one your account is on.

Tenant	Server URL
Production (US)	`https://mcp.trint.com/mcp`
Production (EU)	`https://mcp.eu.trint.com/mcp`

In Claude: add a custom connector, paste the server URL for your region, and complete the Trint sign-in when prompted. Your assistant will then list the Trint tools below.

How authentication works

Unauthenticated requests receive 401 Unauthorized with a WWW-Authenticate: Bearer header pointing at the protected-resource metadata endpoint (/.well-known/oauth-protected-resource).
The client discovers the Trint authorization server from that metadata and runs a standard OAuth flow.
The resulting access token is sent on each request and forwarded to Trint's API. The server never stores your credentials.

Available tools

Reading content

Tool	What it does
`search`	Search your files (transcripts, narratives, captions). Filter by workspace, folder, file type, access role, creator, speaker, transcript ID, and created/updated date ranges; sort and paginate.
`get_file_content`	Return a file's content, optionally scoped to a start/end time range (in seconds).
`get_transcript`	Return a file's transcript, optionally scoped to a start/end time range.
`get_file_metadata`	Fetch metadata for a single file.
`get_folder_metadata`	Fetch metadata for a folder.
`get_workspace_metadata`	Fetch a single workspace by ID.
`list_workspaces`	List the workspaces you can access.
`get_times_from_quotes`	Map an array of quote strings to their positions in a file — each returned with its timecode, speaker, and a deep link into the Trint editor.
`get_suggestions`	Read the current AI suggestion outputs for a content-suggestions record.

Playback (interactive)

These tools return an interactive HTML5 media player, rendered inline by hosts that support MCP Apps UI (the media streams directly from Trint / S3).

Tool	What it does
`render_file_player`	Play a file's media export (video if present, otherwise audio).
`render_quote_player`	Play a specific quote from a file.

HLS playback is Safari-only

HLS (.m3u8) sources play natively in Safari-based hosts only.

Writing (AI Live only)

Tool	What it does
`save_suggestions`	Persist AI-extracted quote suggestions during an AI Live session. Only used when an AI Live `contentSuggestionsId` is supplied; not used for general lookups.

Permissions and data

The assistant can only see files and workspaces you already have access to.
Read tools do not modify your Trint data. save_suggestions writes only within an active AI Live session you initiated.
The server does not collect data beyond what each tool needs to answer a request, and does not access the assistant's memory, chat history, or your other files.

See the Trint Platform Privacy Policy for details.

Support

Questions or issues? Email support@trint.com.