accessoticketing-mcp

An MCP server for accesso Passport mobile ticket links — the URL a venue emails you after a purchase. Reads the order and every admission on it, the scannable barcodes, and Google Wallet save links.

accesso powers ticketing for a lot of theme parks, water parks, camps, museums and festivals, so one server covers all of them: the merchant is encoded in the link.

Developed and maintained by AI (Claude Code). Use at your own discretion.

What it needs

No login. These links carry their own credential — the oToken in the query string is what authorises the read. There is no account, no password, no browser session, and no bot wall.

Related MCP server: tokenless-zendesk-mcp

Install

npx -y @chrischall/accessoticketing-mcp

// claude_desktop_config.json / .mcp.json
{
  "mcpServers": {
    "accessoticketing": {
      "command": "npx",
      "args": ["-y", "@chrischall/accessoticketing-mcp"],
      "env": { "ACCESSO_TICKET_URL": "https://media-engine.na3.accessoticketing.com/tickets/v1/..." }
    }
  }
}

ACCESSO_TICKET_URL is optional — it just saves passing url on every call. The server starts fine without it and reports the missing link on first use.

Tools

Nothing here mutates an order. Every tool is read-only except accesso_save_barcodes, which writes PNG files to disk (never overwriting one) on a local install, or to the host's blob store behind a signed, expiring link when hosted, unless asked for inline images.

> what do we have booked tomorrow?
9:00 AM   Tiny Trekkers Full-Day (Ages 4-7)   — one participant
9:00 AM   Whitewater Kayak Camp I: Intro      — two participants
all-day   Lunch / Snack / Early Drop-Off      — nine more admissions

Upgrading

next release

accesso_save_barcodes no longer writes to the server's disk on a hosted deployment (mcp-host). Local installs are unchanged: same files, same { outputDir, saved: [{ index, participant, packageName, path }] } result.

• Hosted, inline: false (or a server with a writable disk AND a blob store, called without inline): returns { order, note, saved: [{ index, participant, packageName, barcodeText, url, expiresAt }] } — one signed download link per barcode, expiring at expiresAt. There is no outputDir and no path. Previously a hosted server without ACCESSO_NO_FILE_OUTPUT wrote the PNGs to the runner's disk and returned paths nobody could open.

• Inline results (returned, tickets, …) gain linked: [{ …, url, expiresAt }] and linkNote when a blob store is available: barcodes past the inline size cap now come back as links instead of being dropped, so omittedForSize / omittedIndexes only appear when there is no blob store. With no blob store and inline: false (nowhere to save), the inline result carries a note saying so.

• When not a single barcode fits the inline cap and there is no link or disk route, the tool now fails with a too_large error instead of returning an empty image list.

• On a hosted deployment ACCESSO_NO_FILE_OUTPUT is no longer needed, and accesso_healthcheck reports filesPersist: false there.

Shell-out skill

Prefer a shell? skills/accesso-tickets/ does the same reading with curl and a dependency-free parser, no server required. It ships with the package.

Notes on the data

• date / time are the merchant's local wall-clock strings with no timezone attached. They are passed through verbatim rather than converted.

• Merchant-specific detail rows (Guest Number, Web Sales ID, …) are harvested generically into details, so a venue this was never tested against still works.

• An expired link returns HTTP 200, not an error status — the server detects the body and says the link is expired rather than reporting zero tickets.

Development

npm install
npm run build
npm test
npm run test:coverage   # 100% thresholds, enforced in CI

See docs/ACCESSO-API.md for the captured request/response shapes and CLAUDE.md for repo conventions.

License

MIT