webcal-mcp

A small read-only Model Context Protocol server that exposes one or more remote iCalendar feeds (the webcal:// or .ics kind) as queryable tools for an LLM agent.

Install

pipx install webcal-mcp

Or from a checkout:

pipx install .

This puts a webcal-mcp script on your PATH. Python 3.10 or later is required.

Configure

Create ~/.config/webcal-mcp/config.toml (or point WEBCAL_MCP_CONFIG at any other path). See config.example.toml:

default_ttl_seconds = 900

[calendars.personal]
url = "webcal://example.com/personal.ics"
description = "Personal calendar"

[calendars.work]
url = "https://example.com/work.ics"
description = "Work calendar"

webcal:// and webcals:// URLs are normalized to https://.

Local calendars via EventKit (macOS)

On macOS, any calendar already set up in Calendar.app — iCloud, local, subscribed .ics, CalDAV, Google — can be read directly through Apple's EventKit framework. No webcal:// URL hunting required:

[calendars.personal]
source = "eventkit"
identifier = "Personal"   # display name from Calendar.app, or its UUID
description = "iCloud personal calendar"

identifier matches the calendar's display name first, then falls back to its UUID. Run webcal-mcp list-eventkit to see what's available (this dumps title uuid [source] for every calendar).

First use triggers a one-time TCC prompt for Calendar access; the grant is bound to the binary that invokes EventKit (the Python interpreter behind webcal-mcp), so re-granting may be needed if you reinstall. The PyObjC dependency installs automatically on macOS only.

Run

webcal-mcp

The server speaks MCP over stdio. Wire it into an MCP-aware client (Claude Desktop, Claude Code, etc.) by pointing at the webcal-mcp script. For example, for Claude Code:

claude mcp add webcal -- webcal-mcp

Tools

Recurring events are expanded within the requested window. Responses are cached in memory with a per-calendar TTL; revalidation uses HTTP ETag and Last-Modified.

Read-only

Calendars are read-only today. The CalendarSource abstraction reserves a writable capability flag so future write-capable backends can be added without changing the tool surface.