mta-mcp

When to use something else

If your question is "is the L running right now," use mta.info or any transit app. They're free, official, and faster than this.

This server is for a narrower question: a future date, a specific station, and a yes or no you'd put in an email to everyone coming. That's tedious to check by hand and easy to get wrong.

Simple first: which station did you mean?

Harder: "affected" can mean better

This is the case that trips up a simple implementation.

How to use it

Once it's installed (see Running it), you don't call the tools yourself. You ask Claude a question in plain English, and Claude picks a tool and fills in the arguments. It works best when the question names a line, a station, and a date.

Try asking

Illustrative Claude chooses the tool, so the call it makes may differ a little from the one shown. The tool names and parameters are the current ones.

If a station name matches more than one station, the tool returns the candidates instead of guessing, and Claude should ask you which one you meant.

Worked example: a travel note for an event

check_route_on_date({
  "route_id": "6",
  "date": "2026-09-19",
  "stop_id": "628",
  "include_accessibility": true
})

{
  "route_id": "6",
  "date": "2026-09-19",
  "disrupted": false,
  "station": {
    "stop_id": "628",
    "stop_name": "68 St-Hunter College",
    "routes": ["4", "6", "6X"],
    "accessibility": { "status": "fully_accessible", "mta_notes": null }
  },
  "station_level_detail": true,
  "alert_count": 1,
  "alerts": [{
    "alert_type": "Planned - Part Suspended",
    "effect": "reduced",
    "header_text": "No [6] between Hunts Point Av and 125 St",
    "human_readable_active_period": "Sep 18 - Oct 19, Fri 9:30 PM to Mon 5:00 AM",
    "affected_stops": [
      { "stop_id": "614", "stop_name": "Longwood Av" },
      …five more Bronx stations…
    ],
    "affects_this_station": false,
    "relevant_to_station": false
  }],
  "accessibility_outages": {
    "outage_count": 0,
    "outages": [],
    "note": "… No outage is listed here. A missing outage is not proof
      the station is usable: the feed can lag, and an outage MTA hasn't
      entered won't show. Confirm at
      https://www.mta.info/elevator-escalator-status before travel."
  },
  "disclaimer": "Unofficial. BetaNYC is not the MTA …"
}

Worked example: elevators and the way around

get_accessibility_outages({
  "stop_id": "414",
  "route_id": "4",
  "date": "2026-09-26"
})

{
  "date": "2026-09-26",
  "query": "161 St-Yankee Stadium",
  "stop_id": "414",
  "route_id": "4",
  "station_accessibility": { "status": "fully_accessible", "mta_notes": null },
  "outage_count": 5,
  "date_caveats": [],
  "outages": [
    …four more elevators…
    {
      "station": "161 St-Yankee Stadium",
      "trainno": "B/D/4",
      "equipment": "EL133",
      "equipmenttype": "EL",
      "serving": "mezzanine to Manhattan-bound 4 platform",
      "ADA": "Y",
      "outagedate": "11/06/2024 09:00:00 AM",
      "estimatedreturntoservice": "09/30/2026 11:59:00 PM",
      "reason": "Capital Replacement",
      "matched_by": "name",
      "match_note": "MTA's equipment inventory places EL133 at 161 St-Yankee Stadium (D11), another station in this complex. The outage feed's station name fits this station too, so it is listed here.",
      "inventory": {
        "stop_ids": ["D11"],
        "ada_compliant": "YES",
        "redundant_elevator": "-",
        "alternative_route": "To access street level: Take a Manhattan-bound
          4 to 125 St. Use elevators to transfer to a Bronx-bound 6. Exit at
          Hunts Point Av and transfer to a Washington Heights-bound Bx6-SBS
          bus to E 161 St and River Av. …"
      }
    }
  ]
}

Every parameter and every field in every answer is in the tool reference. The elevator, escalator, and ADA data, and where it falls short, are in docs/accessibility.md.

What it cannot do

Questions this server can't answer:

Why this is not on npm

BetaNYC publishes seven MCP servers to npm. This one is intentionally not one of them, because of MTA's data terms.

MTA's data feed terms, term 1:

"In developing your app, you will provide that the MTA data feed is available to others only from a non-MTA server. Accordingly, you will download and store the MTA data feed on a non-MTA server which users of your app will access in order to obtain data. MTA prohibits the development of an app that would make the data available to others directly from MTA's server(s)." (Emphasis ours.)

A published package would have every user's machine fetch straight from api-endpoint.mta.info, with no server of ours in between. You could argue that each user is just fetching for themselves. We don't think that holds up, because the clause defines the non-MTA server as the one "which users of your app will access in order to obtain data." That covers those same users.

So package.json sets "private": true, and an accidental publish fails.

Running it

1. Clone and build

You need Node 18 or newer, and no API key.

git clone https://github.com/BetaNYC/mta-mcp.git
cd mta-mcp
npm install    # also builds

2. Point your AI app at the build

For Claude Code, in a project's .mcp.json:

{
  "mcpServers": {
    "mta-mcp": {
      "command": "node",
      "args": ["${HOME}/Code/mta-mcp/dist/index.js"]
    }
  }
}

For Claude Desktop, in claude_desktop_config.json. Claude Desktop doesn't expand ${HOME}, so use the full path to your copy:

{
  "mcpServers": {
    "mta-mcp": {
      "command": "node",
      "args": ["/absolute/path/to/mta-mcp/dist/index.js"]
    }
  }
}

3. Ask a question

Restart the app and ask something from Try asking, like "Is the 6 train running normally to 68 St–Hunter College on Saturday?"

Using Claude Code and want to keep conversations lighter? The same tools also come as a skill, with no MCP server to run. See skills/README.md.

Being gentle with a public feed

MTA publishes no rate limit, no refresh cadence, and no cache headers, so the server sets its own limits. Six are on by default: a 60-second response cache, single-flight de-duplication, a process-wide one-second floor between requests, bounded retry that honors Retry-After, a 15-second timeout, and a User-Agent that identifies the project.

There's no background polling, prefetch, or warm-up fetch. It fetches only when a tool is called. The test suite makes no network requests, so continuous integration never touches MTA.

What we could not verify

• MTA's documentation and the live feed disagree in at least one place. The docs say a status rank is appended to every alert id. We found it on 1 of 150 entities. We don't know if that's a partial rollout or a documentation error, so we use a different field.
• The refresh cadence is undocumented. MTA doesn't publish it, and responses carry no cache headers. Our one-second minimum between requests is a cautious guess.
• We only inspected the subway feed. MTA says bus, Long Island Rail Road, and Metro-North alerts follow the same conventions. The status rank above is one place they don't.
• Six documented fields never appeared in any live sample we pulled, so we don't use them.

Other NYC subway MCP servers

For most subway questions, one of these is a better fit. Several are further along than ours, and all of them do live arrivals, which this server doesn't. If your question is "when is the next train," start with one of these.

What this server adds: it takes a date. The servers above answer about right now. Of the four, none documents a date parameter in its README, and none says it separates service that got worse from service that got better at a station. That's the gap this server fills.

Community

Where to take your question:

BetaBuilders

In BetaNYC's own words, "a BetaBuilder is a New Yorker who sustains the city's civic infrastructure." The BetaNYC Discord is where we plan classes, fellowships, and open data work, post projects and job openings, and talk through tools like this one while they're still rough.

There are several ways to join:

• Show up. Attending a class or registering for a flagship event earns 90 days of access at no cost.
• Give monthly. $5 a month funds School of Data, Mapping for Equity, Open Data Week, CityCamp NYC, and the weekly newsletter.
• Serve. Fellows, alumni, Open Data Ambassadors, volunteers, working-group members, and community partners are invited directly.
• Scholarships. If cost is the obstacle, say so. BetaNYC's Code of Conduct commits to prioritizing access for people traditionally excluded from civic work.

Learn more about BetaBuilders →

Credits

Data from the MTA service alerts feed, free to use under MTA's terms. Station names from MTA's published static GTFS. MTA's own implementation notes live at nymta/gtfs-documentation, alongside MTA's developer pages.

About BetaNYC

BetaNYC builds and maintains this project. We're New York's civic technology and open data community, and we work to improve lives in New York through civic design, technology, and data.

These tools are free and open source. If you want to help keep the work going, donating is the most direct way.

Link note. Every outbound link on this page was HTTP-checked on 2026-09-21. The mta.info and claude.ai links return 403 to a command-line request because of bot protection, not because they are dead; both were confirmed in a real browser. MTA redirects www.mta.info to mta.info, so the bare form is used here.

AI disclosure. This server and this page were largely written with Claude. Every response shape shown here comes from a real call, which is what the badges mark.