run.events' Model Context Protocol (MCP)

Every run.events event now has its own MCP server

Every event published on run.events is now reachable over the Model Context Protocol (MCP). Any MCP-capable AI application, whether Claude, an IDE assistant, a company's internal AI assistant, or an agent you build yourself, can connect to an event and read its live data: event details, the agenda, sessions, speakers, ticket types with current prices, and sponsors and exhibitors.

Article Image Alt

There is nothing to switch on. The endpoint exists for every event, and it answers with the same data the event itself runs on. No export, no copy, no synchronisation.

What MCP is

MCP is an open standard that lets AI assistants call external tools and read external data in a uniform way. An application that supports it can be pointed at a server, ask which tools are available, and call them. Most current assistants and development tools support it, so a connection set up once works across all of them.

The address

Each event has one address, built from its slug:

https://mcp.runevents.net/{event-slug}

An event with the slug nordic-cloud-summit-2026 is reachable at https://mcp.runevents.net/nordic-cloud-summit-2026. Opening that address in a browser shows a short page with the event name, dates and connection details. An MCP client connecting to the same address gets the protocol endpoint.

Organisers who want the endpoint under their own name can use a custom domain, for example https://mcp.your-event.com. The hostname then identifies the event and no path is needed. Contact support to have a domain registered for your event.

What a connected assistant can do

Nine tools are available at every endpoint:

Tool Returns
get_event Name, dates, description, venues, online or on-site mode, branding, links
get_agenda The full schedule: sessions and breaks with rooms, times, time zone, speakers and labels
search_agenda Free-text search across the agenda
get_sessions All sessions with abstracts, rooms, times, speakers and their biographies, tracks and materials
search_sessions Free-text search across sessions
get_speakers All speakers with names, taglines, companies, biographies and labels
search_speakers Free-text search across speakers
get_ticket_types Ticket types with standard and current prices, time-based tiers, quantity discounts, add-ons and availability
get_sponsors Sponsors, exhibitors and partners with descriptions, logos, levels, booths, team members, products and downloadable assets

Search queries need at least four characters. All tools are read-only. Nothing connected over MCP can change your event.

In practice this means an attendee's assistant can answer "which sessions on Thursday afternoon cover identity management?", a sponsor's internal agent can pull the current exhibitor list into a briefing, and an organiser can check ticket prices from inside their own AI tooling.

Document outlining the ECS 2027 event details, including dates, event description, and instructions for accessing event metadata via a Model Context Protocol (MCP) server. It provides a public interface for querying agenda, speakers, ticket types, and sponsors with API key authentication.

Access and permissions

Event details returned by get_event are public, so that tool works without credentials. Every other tool requires an event API key. It is the same key you already use for the run.events public API, issued under "Event Configuration" -> "API Keys", and the same permission claims apply:

  • agenda and sessions require agenda access
  • speakers require content access
  • ticket types require ticket access
  • sponsors require partner access

The MCP server keeps no copy of your keys and has no permission model of its own. It forwards the key to the run.events API on each call, and the API decides. A key without the required permission receives a plain error message saying so. A key restricted for the public API is restricted for MCP in exactly the same way.

Connecting a client

Add the event endpoint to your MCP client and send the API key as a bearer token:

Endpoint:  https://mcp.runevents.net/{event-slug}
Header:    Authorization: Bearer <event-api-key>

Clients that cannot set an Authorization header can send the key in an ApiKey header instead. Once connected, the client lists the nine tools and can call them. The transport is streamable HTTP, the current MCP standard, so no local process or bridge is needed.

What it does not do today

The current release exposes data. It does not accept changes to an event over MCP, and it does not answer free-form questions itself. When you ask a connected assistant something, the assistant reads the data through these tools and forms the answer. Conversational access to an event's own AI agents over MCP is the next step and will be announced separately.

Where to start

  1. Pick an event and note its slug from the event URL.
  2. Create an API key under "Event Configuration" -> "API Keys" with the permissions you want to expose.
  3. Point your MCP client at https://mcp.runevents.net/{event-slug} with that key.

Questions and feedback are welcome in the community forum.