eesti.ai events MCP server

A remote, read-only MCP server that exposes events from the Estonian eesti.ai initiative (practical AI workshops and meetups) to MCP clients such as Claude Desktop, Claude Code, Cursor and ChatGPT. It wraps the public vportal.ee search API. Version 1 has no authentication.

It runs as a stateless Cloudflare Worker (free tier, no Durable Objects), and it can also be self-hosted in Docker if you do not want to use Cloudflare.

Tools

search_events

Search events by keyword and time window. Input: { keyword?, dateRelative?: "upcoming" | "past", page?, langcode?: "et" | "en" } Defaults: dateRelative upcoming, page 1, langcode et. Returns { events, total, page, pageSize, hasMore }.

upcoming_events

The next upcoming events, sorted by start time. Input: { limit?, langcode?: "et" | "en" } (limit default 10).

get_event

A single event by its canonical id. Input: { id }. Searches the upcoming then past windows and matches by id.

Each tool returns both a structured JSON payload and a short text summary.

Requirements

• Node.js 20 or newer (for local development), or

• Docker (for the self-hosted container path).

The endpoint speaks MCP over streamable HTTP. The URL is the server root, for example http://localhost:8787/.

Setup: run it locally with Node

Step by step:

1. Clone the repository and enter it.

   git clone <repo-url> eesti-ai-events-mcp
   cd eesti-ai-events-mcp

2. Install dependencies.

   npm install

3. Run the unit tests to confirm everything works.

   npm test

4. Start the server. This runs the Worker locally with workerd, no Cloudflare account needed.

   npm run dev

   Wrangler prints the local URL, normally http://localhost:8787.

5. Smoke-test it with curl.

   curl -s -X POST http://localhost:8787/ \
     -H 'Content-Type: application/json' \
     -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' | jq '.result.tools[].name'

   You should see search_events, upcoming_events, get_event.

Setup: run it in Docker (self-hosted)

This path needs only Docker. It runs the same Worker locally inside the container, so there is no Cloudflare account and no monthly cost.

1. Build and start the container.

   docker compose up --build

2. The MCP endpoint is now at http://localhost:8787/. Test it with the same curl command as above.

3. Stop it with Ctrl-C, or run it in the background with docker compose up -d and stop it later with docker compose down.

To change the host port, edit the ports mapping in docker-compose.yml, for example "9000:8787" to serve on port 9000.

Note: the container runs wrangler dev, which is a development server. It is fine for personal and small-team self-hosting. For a hardened public deployment, prefer the Cloudflare deploy path below.

Connect an MCP client

Point your client at the server URL. Use http://localhost:8787/ for local or Docker, or your deployed https://...workers.dev/ URL for the Cloudflare path.

Claude Code

claude mcp add --transport http eesti-ai http://localhost:8787/

Then list tools with /mcp inside Claude Code.

Cursor

Add this to .cursor/mcp.json (project) or ~/.cursor/mcp.json (global):

{
  "mcpServers": {
    "eesti-ai": {
      "url": "http://localhost:8787/"
    }
  }
}

Claude Desktop

Claude Desktop connects to remote HTTP servers through the mcp-remote bridge. Open Settings, Developer, Edit Config to open claude_desktop_config.json.

This file usually already has other settings in it. Do not paste over the whole file. Add only the mcpServers block. If you already have an mcpServers block, add the eesti-ai entry inside it and leave the rest alone.

The part to add:

"mcpServers": {
  "eesti-ai": {
    "command": "npx",
    "args": ["mcp-remote", "http://localhost:8787/"]
  }
}

For context, a full config file with that block merged in looks something like this. Your other keys will differ, so keep yours and only add mcpServers:

Warning: do not copy the example below. It is for illustration only, to show where the mcpServers block sits among other keys. The preferences, coworkUserFilesPath, and other values are placeholders. Copying it will overwrite your real settings. Only add the mcpServers block shown above to your own existing file.

{
  "preferences": {
    "remoteToolsDeviceName": "your-device-name",
    "coworkWebSearchEnabled": true,
    "coworkScheduledTasksEnabled": true,
    "ccdScheduledTasksEnabled": true
  },
  "coworkUserFilesPath": "/Users/you/Documents/Claude",
  "mcpServers": {
    "eesti-ai": {
      "command": "npx",
      "args": ["mcp-remote", "http://localhost:8787/"]
    }
  }
}

Save the file and restart Claude Desktop. The eesti.ai tools appear in the tools menu.

MCP Inspector (for testing)

npx @modelcontextprotocol/inspector

In the Inspector, choose transport "Streamable HTTP", enter the server URL, and exercise the three tools.

Deploy to Cloudflare (optional)

1. Log in once.

   npx wrangler login

2. Deploy.

   npm run deploy

   Wrangler prints the public https://eesti-ai-events-mcp.<your-subdomain>.workers.dev URL. Use that URL in the client configuration above. A custom domain is optional and can be added later in the Cloudflare dashboard.

Notes

Version 1 is a stateless Worker on the Cloudflare free tier. It does not use Durable Objects, so there is no Workers Paid plan cost. The English language code is accepted but returns no events today, because the upstream has no English data yet.