Brussels Live

What it is

Brussels Live has a server for the Model Context Protocol (MCP), the open standard AI assistants use to work with outside sources. Once you have connected it, your assistant can search the broadcasts, read transcripts and summaries, and check the agenda whenever your question calls for it.

It works with everything that is on the site: plenary sittings and committee meetings of the European Parliament, the Commission's midday briefing, and press conferences and Council meetings broadcast live on EbS. Live broadcasts are readable while they are still going on.

How to connect

Paste the server address into your assistant as a custom connector or MCP server. Menu names change now and then; look for "connectors" or "MCP".

Claude (claude.ai, desktop and mobile)

Open Settings → Connectors and choose Add custom connector.
Name it Brussels Live and paste the server address. Leave the advanced settings empty: no login is needed.
In a chat, switch the connector on in the tools menu, then just ask your question.

On Team and Enterprise plans an owner first adds the connector in the organisation settings.

Claude Code

Run this once in your terminal:

claude mcp add --transport http brussels-live https://brusselslive.eu/api/eu/mcp

ChatGPT

Open Settings → Apps & Connectors, and under Advanced settings switch on developer mode.
Create a connector, paste the server address and choose no authentication.
In a chat, select the connector under the + menu.

Which plans can add their own connectors is up to OpenAI.

Mistral Vibe (Work)

In Vibe, go to Work → Connectors, choose + Add Connector and the Custom MCP Connector tab.
Name: brusselslive (no spaces). Server URL: the address above. No authentication.
Click Connect, then switch the connector on in a chat.

Cursor, VS Code and other MCP clients

Add the server as a remote server (Streamable HTTP), for example:

{
  "mcpServers": {
    "brussels-live": {
      "url": "https://brusselslive.eu/api/eu/mcp"
    }
  }
}

The server works without sessions or login, so no extra settings are needed.

What the assistant can do

The server offers six tools. You do not have to name them: the assistant picks them itself.

search_transcripts: Searches what was said, in any EU language, with filters for the kind of broadcast, the dates and the language actually spoken.“Which MEPs talked about Erasmus+ this month?”
list_broadcasts: Lists broadcasts by kind, date, status or committee.“Which LIBE committee meetings were there in October?”
get_broadcast: Shows one broadcast: the agenda with its debates, the speakers and the languages spoken.“What was on the agenda of yesterday's plenary?”
get_transcript: Reads the transcript in your language, the original or both, per debate or from a given time, with a link to every turn.“What did the Commissioner say about the budget at the start of the debate?”
get_summaries: Gives the AI summaries per debate: topics, commitments, critical questions and notable moments, with quotes checked against the transcript.“Summarise the debate on the rule of law in Malta.”
get_agenda: Shows what is coming up in the next seven days and what we transcribe live.“What is on the agenda tomorrow, and can I follow it live?”

In your language

Ask in any of the 24 official EU languages. The assistant reads the transcripts in that language and links to the site in that language. Everything is translated into English first, so for the widest search it may also search in English.

AI and sources

Transcripts come from speech recognition of the original sound in the room (never the interpreters), translations are made by machine and summaries are written by AI. Nothing is checked by an editor first.
The original language is authoritative. The transcripts are not an official record: the institution's own record is what counts.
Your assistant adds its own layer of AI: what it writes about the transcripts is its own work. Ask it for the links, and check the original and the video before you quote.

How we meet the EU AI Act

Privacy and limits

Your conversation stays between you and your AI provider, under its terms. Our server only sees what the assistant asks it, such as a search term, and does not keep it.
Up to 120 requests and 30 searches per minute per IP address. Web versions of assistants share their provider's addresses: if you see "too many requests", wait a minute.
Searching without dates covers the last 90 days; with dates up to a year at a time.

Brussels Live in your AI assistant