Introducing the VIDTREO MCP: your recordings, inside your AI assistant

TL;DR

• The VIDTREO MCP is live. Connect Claude, Codex or any MCP client to your VIDTREO workspace with one URL: https://mcp.vidtreo.com/mcp.
• Ask in plain words. Find a recording, read its transcript, play it, record a new one, share it or analyze it with Intelligence. The assistant has 37 tools, and you never call one yourself.
• Record and watch in the chat. Claude on the web and Claude Desktop show the VIDTREO recorder and player inside the conversation.
• Your permissions, not more. You sign in with your VIDTREO account and choose what to grant, and your role is checked on every call. There’s no API key to copy and no extra charge for the MCP.

Your work is in the chat. Your recordings are in another tab

More and more work starts in an AI assistant: the follow-up email after a sales call, the notes before an interview, the bug report after a support call.

The recording all of that depends on lives somewhere else. You open the dashboard, find the video, scrub to the right minute, copy what was said and paste it back into the chat.

Last week, Intelligence started turning every recording into answers with evidence. The VIDTREO MCP brings those answers, and the recordings themselves, to the conversation where you already work.

What the VIDTREO MCP is

MCP, the Model Context Protocol, is the open standard AI assistants use to connect to other services. The VIDTREO MCP is a remote MCP server for your VIDTREO workspace.

It holds no data of its own. Every request goes to the VIDTREO Edge API with your sign-in, where your role and the permissions you granted are checked.

It works with Claude on the web, Claude Desktop, Claude Code, Codex and any client that supports remote MCP servers with OAuth.

Ask for what you need

You don’t learn tool names. You ask, and the assistant picks the tools:

You ask	What happens
”Show me the videos recorded in the Sales environment this month”	Searches videos and audios by title, environment, status and date
”Summarize what the candidate said about their last job”	Reads the transcript, with timestamps, and answers from it
”Play the onboarding video”	Shows the VIDTREO player inside the conversation
”Record an audio note called Weekly update”	Shows the VIDTREO recorder, and the recording uploads to your workspace
”Make the demo public and invite [email protected]”	Changes the visibility and emails a personal link, which it can revoke later
”Create an environment for interviews in Full HD with AI transcription”	Creates and updates environments and sets your default one
”Analyze this call with the Sales discovery lens”	Starts an Intelligence analysis and reads the result
”Did the customer mention a budget? Cite the moment”	Answers from the video’s analysis, with the moments that support it

One request can chain several tools. Try it with a sample recording: pick a request, open any tool call to see what the assistant sent and got back, and play or record right in the conversation.

AI assistantVIDTREO MCPSample

Your VIDTREO workspace is connected

Pick a request or type your own. The assistant chooses the tools.

Find the Acme discovery call. What did they push back on?Play the Acme discovery callRecord an audio noteHelp me create a lens to screen SDR candidates

Record and watch without leaving the chat

The VIDTREO MCP includes two MCP Apps, interactive pages the assistant shows inside the conversation: the recorder and the player.

The recorder is the VIDTREO Recorder. Ask “Record an audio note called Weekly update” and it appears in the chat. When you stop, the recording uploads to your workspace with the title you chose. The assistant never invents a title: if you didn’t give one, it asks.

Mode	Claude (web)	Claude Desktop
Audio	In the conversation	In the conversation
Camera	In the conversation, when the browser grants camera access	Continues in VIDTREO
Screen, or screen and camera	Continues in VIDTREO	Continues in VIDTREO

“Continues in VIDTREO” means the conversation shows a Record in Vidtreo button. The VIDTREO Dashboard opens in the right environment, and once you finish, the chat shows the recording as ready, with your title. Apps inside an assistant can’t capture your screen, so screen recordings always take this path.

A recording made from the assistant is like any other. It uses the environment’s quality, audio and transcription settings and fires its webhooks, and if a lens has auto-analyze on in that environment, it gets analyzed.

The player is the VIDTREO Player. Ask “Play yesterday’s customer call” and the video or audio streams in the chat, straight from VIDTREO storage. The player is for you: to answer questions about a recording, the assistant reads its transcript or its Intelligence analysis.

In Claude Code and other clients without MCP Apps, the assistant gives you a link to the recorder or the recording in the VIDTREO Dashboard instead.

Intelligence, from the conversation

Everything you do with Intelligence in the dashboard, you can also do from the chat:

• Analyze. “Analyze the Acme call with the Sales discovery lens.” The assistant starts the analysis, waits for it and answers from the result: summary, data points, chapters and scorecard.
• Ask. “Did they mention a budget? When?” The answer comes from the video’s latest analysis, with citations to the moments that prove it.
• Design a lens. “Help me create a lens to screen SDR candidates.” The assistant interviews you in short rounds, using what it already knows about your company, drafts the whole lens and shows it to you. Then it asks whether to publish it or keep a draft, and where to activate it.
• Keep it current. Activate a lens in an environment, change it and publish a new version. Every change is a new version, and earlier ones are kept.

Analyses started from the assistant are the same analyses you run in the dashboard: they appear there, cost the same and fire the intelligence.lens_analysis.completed webhook. Activating a lens from the chat leaves auto-analyze off until you ask for it.

Connect in a minute

Every client uses the same URL. There’s no API key: the client opens the VIDTREO sign-in page, you approve the permissions and it stays connected.

Claude (web and desktop). Open Settings → Connectors → Add custom connector, enter https://mcp.vidtreo.com/mcp, then click Connect and sign in. Add it once and both apps can use it.

Claude Code.

claude mcp add --transport http vidtreo https://mcp.vidtreo.com/mcp

Then run /mcp, select vidtreo and choose Authenticate.

Codex.

codex mcp add vidtreo --url https://mcp.vidtreo.com/mcp
codex mcp login vidtreo

Any other MCP client that supports remote servers over Streamable HTTP with OAuth connects with the same URL. It discovers the rest from the server: dynamic client registration, authorization code with PKCE and refresh tokens.

On Claude Team and Enterprise plans, an organization Owner may need to add the connector first.

Your permissions, not more

Connecting an assistant to your recordings should never mean handing it the keys.

• You choose the scopes. The consent screen lists them in plain words, from View videos to Delete recordings and Run intelligence on videos. Grant only what you want the assistant to do.
• Your role always applies. Every call is checked against your role in the workspace and the permissions you granted. A Viewer who connects the MCP is still a Viewer.
• It asks before naming, deleting or spending. Titles and names come from you. Deleting, revoking a link or changing who can watch are marked as destructive, so your client can ask for your approval. Analyses only start when you ask, after the assistant mentions the cost.
• Media links stay private. The recorder gets a short-lived upload token for one environment, never your sign-in token. Playback links are temporary and only reach the player, so the assistant never sees them.
• The server stores nothing. It verifies the signature, issuer and audience of every token and forwards the call to the VIDTREO Edge API, where everything is checked again.
• Revoke access any time. Disconnect from your client, or open Company → Connected apps in the VIDTREO Dashboard.

Your own S3 storage and webhook endpoints stay in the dashboard. The MCP never reads or changes them.

What it costs

The MCP itself costs nothing. It uses your workspace’s prepaid credits the same way the dashboard does:

Action	Charged as
Searching, reading transcripts and results, sharing, configuring	Free
Recording from the assistant	Any other recording in that environment
Analyzing a video	One Intelligence analysis, $0.035 per analyzed minute
Turning on auto-analyze	One analysis for every new recording in that environment, per lens

What’s not in the MCP yet

• Analyses run on videos. Analyze audio recordings from the dashboard. Their transcripts are available to the assistant.
• Ask works on one video at a time. Asking across all your recordings is in the dashboard.
• Screen recording continues in VIDTREO. The dashboard records it and the chat picks it up.
• Analyses from the chat use the default inputs: Spanish and English as the spoken language, with no question and no participant names.

Get started

1. Create your account and get $1 in free credit.
2. Add https://mcp.vidtreo.com/mcp to Claude, Claude Code, Codex or your MCP client.
3. Ask “List my VIDTREO environments”, then “Record an audio note called First test”.

Your recordings already have the answers. Now you can ask for them where you work.

Start free — $1 in credit → Read the MCP docs → See how Intelligence works →