Skip to main content
Glama
README.md
# river-level-mcp

Live US river levels for AI assistants: river stage, streamflow and water temperature from USGS stream gauges, with NWS flood stages and river forecasts.

This is a Model Context Protocol (MCP) server. It reads the USGS Water Data API and the NOAA National Water Prediction Service at call time. It needs no API key.

## Tools

| Tool | What it returns |
| --- | --- |
| `find_river_gauges` | Active USGS river and stream gauges by river and place name, state, or latitude and longitude, with each gauge's latest stage and flow. Active gauges first. |
| `get_river_level` | Current river stage (gage height, ft), streamflow (discharge, ft3/s) and water temperature where measured, the measurement time, NWS flood stages, current flood category, feet below flood stage and record crest. Flags readings over 48 hours old. |
| `get_river_history` | Daily mean streamflow or stage for 2 to 366 days, with first, last, low, high, mean and change. |
| `get_flood_forecast` | The current NWS river forecast for a gauge: forecast series, forecast crest and its flood category, latest observation, flood stages and record crest. |

Every result has structured JSON, a one-paragraph summary, `as_of`, `source` (the USGS or NWS gauge page) and `more` (a page on [lakelevelnow.com](https://lakelevelnow.com/?utm_source=mcp&utm_medium=river-level-mcp)).

## What it knows that your model does not

- Today's river level. Gauges report every 5 to 60 minutes and a river can rise 10 feet in a day. A model's training data is months old.
- Whether a river is near flood stage right now, using the NWS action, minor, moderate and major flood stages for that gauge.
- The latest NWS river forecast and its crest, reissued several times a day.
- The new USGS Water Data API. The legacy waterservices.usgs.gov service shuts down on February 22, 2027, so code that calls it will stop working. This server already uses api.waterdata.usgs.gov.

## Install

### Hosted (Claude.ai custom connector, ChatGPT developer mode)

Paste this URL. No token, no sign-up.

```
https://river-level-mcp.lipmichal.workers.dev/mcp
```

### Claude Code

```
claude mcp add --transport http river-level https://river-level-mcp.lipmichal.workers.dev/mcp
```

Or run it locally over stdio:

```
claude mcp add river-level -- npx -y github:theluckystrike/river-level-mcp
```

### Claude Desktop

Add to `claude_desktop_config.json`:

```json
{
  "mcpServers": {
    "river-level": { "command": "npx", "args": ["-y", "github:theluckystrike/river-level-mcp"] }
  }
}
```

### Cursor

Add to `~/.cursor/mcp.json`:

```json
{
  "mcpServers": {
    "river-level": { "url": "https://river-level-mcp.lipmichal.workers.dev/mcp" }
  }
}
```

### VS Code

Add to `.vscode/mcp.json`:

```json
{
  "servers": {
    "river-level": { "type": "http", "url": "https://river-level-mcp.lipmichal.workers.dev/mcp" }
  }
}
```

Node 20 or newer is needed for the local stdio version.

## Example prompts

- "What is the Mississippi River level at St. Louis right now?"
- "Is the Potomac at Little Falls near flood stage?"
- "How much water is flowing in the Colorado River at Austin, and how has it changed over the last 90 days?"
- "When will the river crest at EADM7, and will it reach flood stage?"
- "Which stream gauges are within 10 miles of 38.95, -77.13?"

## Data sources

Full list with fetch dates in [DATA_SOURCES.md](DATA_SOURCES.md).

- River stage, streamflow, water temperature, daily means and site details from the [USGS Water Data API](https://api.waterdata.usgs.gov/) (OGC API collections `latest-continuous`, `daily`, `monitoring-locations`).
- Flood stages, record crests and river forecasts from the [NOAA National Water Prediction Service API](https://api.water.noaa.gov/nwps/v1/docs/).
- The hosted server caches readings at the edge for a few minutes and site details for a day, so repeat calls do not reach USGS or NOAA.

## Limitations

- USGS real-time data are provisional and can be revised. The tools say so.
- Only gauges that are NWS forecast points have flood stages and forecasts. Other gauges still return the USGS reading.
- NWS stage and USGS gage height normally use the same gauge datum, but check the NWS gauge page before relying on a threshold comparison for safety. Follow local NWS warnings for flood decisions.
- The USGS API applies a rate limit to keyless callers. When it answers 429 or is down, the tool returns a clear error instead of a stale or remembered number.
- Lakes and reservoirs are covered by [lake-level-mcp](https://github.com/theluckystrike/lake-level-mcp).

## More

Live lake and reservoir levels for Lake Mead, Lake Powell, Lake Travis and more are on [lakelevelnow.com](https://lakelevelnow.com/?utm_source=mcp&utm_medium=river-level-mcp).

Part of the Zovo MCP collection: https://mcp.zovo.one

## Development

```
npm install
npm test            # builds, then runs the fixture tests against saved USGS and NOAA payloads
node scripts/smoke-stdio.mjs   # live stdio check, one real call per tool
npx wrangler deploy            # Cloudflare Worker, Streamable HTTP at /mcp
```

## License

MIT

TDQS

A4.3/5.0

Scored across 4 tools

Disambiguation5/5

Each tool maps to a distinct temporal/functional role: find_river_gauges for discovery, get_river_level for current conditions, get_river_history for past trends, and get_flood_forecast for predictions. The only minor overlap is that get_river_level and get_flood_forecast both surface flood stages, but the primary purpose of each is clear and an agent can select confidently.

Naming Consistency5/5

All tools follow a clean verb_noun pattern (find_river_gauges, get_river_level, get_river_history, get_flood_forecast) with the river/gauge domain object consistent across names. The find-vs-get distinction is semantically meaningful (search vs. retrieve) rather than arbitrary.

Tool Count5/5

Four tools is well-scoped for a focused USGS gauge data server: discover, current reading, history, and forecast. Each earns its place with no redundant or filler tools.

Completeness4/5

The surface covers the full read lifecycle from gauge discovery through current readings, historical trends, and NWS forecasts, which addresses the common question types. Minor gaps remain, such as comparing multiple gauges in one call or basin-wide summaries, but agents can work around these via repeated queries.

Maintenance

ActivityMaintained
ResponsivenessNo issues