Parks Open, for AI agents

Built to be read by machines as well as people. An assistant can query live campsite availability across BC Parks — by real equipment fit and the site attributes the official sites don’t expose — and hand back a booking link into the official flow. Every answer carries observed_at; nothing is booked on anyone’s behalf.

Add to Claude

Settings → Connectors → Add custom connector → paste https://parksopen.ca/api/mcp. No key is needed; the tools are read-only. Then try one of the prompts below, or the built-in prompt plan_bc_camping_weekend.

Add to ChatGPT

Settings → Connectors → Developer mode → Create with the MCP URL above, or add the OpenAPI description at /openapi.json as an Action (link-out only — there is no checkout here).

Cursor, Windsurf, Claude Code

{
  "mcpServers": {
    "parksopen": {
      "url": "https://parksopen.ca/api/mcp"
    }
  }
}

Claude Code: claude mcp add --transport http parksopen https://parksopen.ca/api/mcp. Inspect with npx @modelcontextprotocol/inspector https://parksopen.ca/api/mcp.

Try asking

“Find a waterfront site that fits a 30 ft trailer near Squamish this long weekend.”
“Is Golden Ears sold out for the weekend of Sep 12? If so, which nearby park has space?”
“When do Alice Lake reservations open for July, and do they allow dogs?”

Tools

parksopen_search_campsites — Search campsites by live availability
parksopen_park_availability_calendar — Park availability calendar (8 weeks)
parksopen_find_open_parks — Find parks with openings (ranked)
parksopen_check_alerts — Closures, fire bans, advisories, nearby wildfires
parksopen_check_weather — Forecast and air quality at a park
parksopen_park_info — Park facts, rules and fees (knowledge base)
parksopen_get_booking_link — Booking link for a park and dates

All tools are read-only, idempotent and closed-world (annotated as such). The MCP server also exposes a park://<system>/<slug> resource per park and a planning prompt. The REST mirror POST /api/tools/<name> takes the same JSON arguments.

Also machine-readable

/llms.txt — what this site is, per system, with every park.
/.well-known/ard.json and /.well-known/agent-card.json — discovery catalogs.
/sitemap.xml — index; one child per system with real lastmod.
Every park page answers first (the live count in the first sentence) and ships Campground + BreadcrumbList JSON-LD.

Limits and privacy

Fair use: about 60 tool calls a minute per client; heavier use — email us. We read public availability only, store no user credentials, never hold or pay for sites, and never train on your prompts (they are not stored). Data is observed, not guaranteed: check observed_at and book on the official site.

Building something on this? Email hello@parksopen.ca.