SWATGenX is an MCP server published by io.github.SWATGenX, hosted at www.swatgenx.com. When last checked on 2026-11-22 it was working: it answers and lists its tools. It declares 13 tools, including search_swat_models, get_model_calibration, query_groundwater, query_pfas, request_model.
SWAT+ watershed models for any river in the lower 48 states, with groundwater and PFAS data
1 connections through 5 shared values. Tap any node.
Squares are this server's values; dots are other servers sharing them. A shared address or publisher is a lead, not proof of a common operator.
Inferred from the host name and headers: what a server runs on, not who runs it.
Search the public SWATGenX example-model catalog (built SWAT+ models for USGS gauges and HUC outlets across the conterminous US). Filter by two-letter `state`, `calibrated_only` (alias `calibrated`), and channel-count bounds. A model counts as calibrated only when the website shows it as calibrated (its calibration tier is advertised); a failed or below-satisfactory run does not. Returns model id (site_no), state, size, the calibration tier, and — when present — the headline calibration daily NSE.
| Parameter | Type | Description |
|---|---|---|
| limit | integer | |
| state | any | |
| calibrated | any | |
| max_channels | any | |
| min_channels | any | |
| calibrated_only | boolean |
{
"name": "search_swat_models",
"title": "Search SWAT+ example models",
"description": "Search the public SWATGenX example-model catalog (built SWAT+ models for USGS gauges and\n HUC outlets across the conterminous US). Filter by two-letter `state`, `calibrated_only`\n (alias `calibrated`), and channel-count bounds. A model counts as calibrated only when the\n website shows it as calibrated (its calibration tier is advertised); a failed or\n below-satisfactory run does not. Returns model id (site_no), state, size, the calibration\n tier, and — when present — the headline calibration daily NSE.",
"inputSchema": {
"type": "object",
"title": "search_swat_modelsArguments",
"properties": {
"limit": {
"type": "integer",
"title": "Limit",
"default": 25
},
"state": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"title": "State",
"default": null
},
"calibrated": {
"anyOf": [
{
"type": "boolean"
},
{
"type": "null"
}
],
"title": "Calibrated",
"default": null
},
"max_channels": {
"anyOf": [
{
"type": "integer"
},
{
"type": "null"
}
],
"title": "Max Channels",
"default": null
},
"min_channels": {
"anyOf": [
{
"type": "integer"
},
{
"type": "null"
}
],
"title": "Min Channels",
"default": null
},
"calibrated_only": {
"type": "boolean",
"title": "Calibrated Only",
"default": false
}
}
},
"annotations": {
"title": "Search SWAT+ example models",
"readOnlyHint": true,
"openWorldHint": false,
"idempotentHint": true,
"destructiveHint": false
}
}Return the published calibration for one example model: method, calibration NSE (daily + monthly), PBIAS, evals used, and window as {fit, scored}: the years the calibration was fitted to and the years its validation was scored on. Validation NSE is included only when the scored years lie outside the fitted years. calibration_status is calibrated_and_validated or calibrated_not_validated only where the model's page shows its calibration badge (calibration_advertised), and uncalibrated otherwise, including when no calibration of the model is published. Where the example model has a recorded calibration run, its numbers are returned with either status: they are that run's scores, and the model you download may not carry that calibration. Returns an error with code=model_not_found if no such model exists at all. site_no is the model's USGS gauge or HUC outlet id.
| Parameter | Type | Description |
|---|---|---|
| site_no * | string |
{
"name": "get_model_calibration",
"title": "Get model calibration results",
"description": "Return the published calibration for one example model: method, calibration NSE (daily + monthly),\n PBIAS, evals used, and window as {fit, scored}: the years the calibration was fitted to and the\n years its validation was scored on. Validation NSE is included only when the scored years lie\n outside the fitted years. calibration_status is calibrated_and_validated or calibrated_not_validated\n only where the model's page shows its calibration badge (calibration_advertised), and uncalibrated\n otherwise, including when no calibration of the model is published. Where the example model has a\n recorded calibration run, its numbers are returned with either status: they are that run's scores,\n and the model you download may not carry that calibration. Returns an error with\n code=model_not_found if no such model exists at all. site_no is the model's USGS gauge or HUC outlet\n id.",
"inputSchema": {
"type": "object",
"title": "get_model_calibrationArguments",
"required": [
"site_no"
],
"properties": {
"site_no": {
"type": "string",
"title": "Site No"
}
}
},
"annotations": {
"title": "Get model calibration results",
"readOnlyHint": true,
"openWorldHint": false,
"idempotentHint": true,
"destructiveHint": false
}
}Query SWATGenX's national groundwater inventory (well records and lithology logs harvested from state agencies across the contiguous U.S.; live totals at https://www.swatgenx.com/api/gw-wells/summary) near a point — the subsurface data used to parameterise MODFLOW. Returns the nearest wells with lithology-interval count and a richness_tier grading what each well EXPOSES: 0 = well record with no water level or yield (many carry no logged lithology), 1 = + hydraulics (static water level / yield / depth), 2 = + screened interval & yield (groundwater-model-ready). The tier is about attribute coverage, not log detail. Per-interval depths live in the full inventory DB (Zenodo), not this index. lat/lon in decimal degrees (EPSG:4326); radius_km up to ~50.
| Parameter | Type | Description |
|---|---|---|
| lat * | number | |
| lon * | number | |
| limit | integer | |
| radius_km | number |
{
"name": "query_groundwater",
"title": "Query national groundwater wells",
"description": "Query SWATGenX's national groundwater inventory (well records and lithology logs harvested\n from state agencies across the contiguous U.S.; live totals at\n https://www.swatgenx.com/api/gw-wells/summary) near a point — the subsurface data used to\n parameterise MODFLOW. Returns the nearest\n wells with lithology-interval count and a richness_tier grading what each well EXPOSES:\n 0 = well record with no water level or yield (many carry no logged lithology),\n 1 = + hydraulics (static water level / yield / depth),\n 2 = + screened interval & yield (groundwater-model-ready). The tier is about attribute\n coverage, not log detail. Per-interval depths live in the full inventory DB (Zenodo), not\n this index. lat/lon in decimal degrees (EPSG:4326); radius_km up to ~50.",
"inputSchema": {
"type": "object",
"title": "query_groundwaterArguments",
"required": [
"lat",
"lon"
],
"properties": {
"lat": {
"type": "number",
"title": "Lat"
},
"lon": {
"type": "number",
"title": "Lon"
},
"limit": {
"type": "integer",
"title": "Limit",
"default": 20
},
"radius_km": {
"type": "number",
"title": "Radius Km",
"default": 5
}
}
},
"annotations": {
"title": "Query national groundwater wells",
"readOnlyHint": true,
"openWorldHint": false,
"idempotentHint": true,
"destructiveHint": false
}
}Query the SWATGenX PFAS monitoring inventory (agency monitoring stations + observations): only sources whose licence allows serving them, the same set as the PFAS data API; unaudited VLM packs are excluded. Filter by two-letter `state` (e.g. 'MI') and/or `analyte` (e.g. 'PFOS', 'PFOA'): with an analyte, only stations that have at least one observation of it are returned, each with its latest recorded value (media, n_samples, n_detect, max_value, latest_value, latest_date, detection band). Units follow the medium: water / groundwater / porewater values are ng/L; sediment, soil and tissue are mass-based (ug/kg-class), so values are comparable across stations only within one `media`. Each answer counts toward the account's monthly allowance of distinct stations (a station already returned this month does not count again) and 200 new stations a day; `metering` in the reply reports the account's count. For research/screening use; not a regulatory data source.
| Parameter | Type | Description |
|---|---|---|
| limit | integer | |
| state | any | |
| analyte | any |
{
"name": "query_pfas",
"title": "Query PFAS monitoring (US or worldwide)",
"description": "Query the SWATGenX PFAS monitoring inventory (agency monitoring stations + observations):\n only sources whose licence allows serving them, the same set as the PFAS data API; unaudited\n VLM packs are excluded. Filter by two-letter `state` (e.g. 'MI') and/or `analyte` (e.g.\n 'PFOS', 'PFOA'): with an analyte, only stations that have at least one observation of it are\n returned, each with its latest recorded value (media, n_samples, n_detect, max_value,\n latest_value, latest_date, detection band). Units follow the medium: water / groundwater /\n porewater values are ng/L; sediment, soil and tissue are mass-based (ug/kg-class), so values\n are comparable across stations only within one `media`. Each answer counts toward the account's\n monthly allowance of distinct stations (a station already returned this month does not count\n again) and 200 new stations a day; `metering` in the reply reports the account's count. For research/screening use; not a regulatory data source.",
"inputSchema": {
"type": "object",
"title": "query_pfasArguments",
"properties": {
"limit": {
"type": "integer",
"title": "Limit",
"default": 25
},
"state": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"title": "State",
"default": null
},
"analyte": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"title": "Analyte",
"default": null
}
}
},
"annotations": {
"title": "Query PFAS monitoring (US or worldwide)",
"readOnlyHint": true,
"openWorldHint": false,
"idempotentHint": true,
"destructiveHint": false
}
}Preview what SWATGenX WOULD build for a CONUS catchment. Takes either a USGS gauge id (usgs_station) or a 12-digit HUC12 outlet (huc12_outlet). Read-only in v1: returns whether the catchment is buildable, whether a model already exists, and the build endpoint (order_endpoint) — it does NOT start a build (that needs an authenticated account at swatgenx.com).
| Parameter | Type | Description |
|---|---|---|
| huc12_outlet | any | |
| usgs_station | any |
{
"name": "request_model",
"title": "Preview a watershed model build",
"description": "Preview what SWATGenX WOULD build for a CONUS catchment. Takes either a USGS gauge id\n (usgs_station) or a 12-digit HUC12 outlet (huc12_outlet). Read-only in v1: returns whether the\n catchment is buildable, whether a model already exists, and the build endpoint (order_endpoint) —\n it does NOT start a build (that needs an authenticated account at swatgenx.com).",
"inputSchema": {
"type": "object",
"title": "request_modelArguments",
"properties": {
"huc12_outlet": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"title": "Huc12 Outlet",
"default": null
},
"usgs_station": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"title": "Usgs Station",
"default": null
}
}
},
"annotations": {
"title": "Preview a watershed model build",
"readOnlyHint": true,
"openWorldHint": false,
"idempotentHint": true,
"destructiveHint": false
}
}Search the SWATGenX I/O reference — model bundle layout, SWAT+ input files, PFAS fate-and-transport inputs, engine run modes and NetCDF output, MODFLOW-6 coupling, calibration artifacts, service limits. Returns ranked articles, each with a matching snippet and its chapter slug. The schemas are transcribed from the engine readers/writers and include the SWATGenX extensions, which stock SWAT+ documentation does not cover.
| Parameter | Type | Description |
|---|---|---|
| limit | integer | |
| query * | string |
{
"name": "search_docs",
"title": "Search the SWATGenX documentation",
"description": "Search the SWATGenX I/O reference — model bundle layout, SWAT+ input files, PFAS\n fate-and-transport inputs, engine run modes and NetCDF output, MODFLOW-6 coupling,\n calibration artifacts, service limits. Returns ranked articles, each with a matching\n snippet and its chapter slug. The schemas are transcribed from the engine readers/writers\n and include the SWATGenX extensions, which stock SWAT+ documentation does not cover.",
"inputSchema": {
"type": "object",
"title": "search_docsArguments",
"required": [
"query"
],
"properties": {
"limit": {
"type": "integer",
"title": "Limit",
"default": 5
},
"query": {
"type": "string",
"title": "Query"
}
}
},
"annotations": {
"title": "Search the SWATGenX documentation",
"readOnlyHint": true,
"openWorldHint": false,
"idempotentHint": true,
"destructiveHint": false
}
}Return one full SWATGenX documentation chapter as markdown, by chapter slug; 'overview' is the index and lists every chapter. Retired slugs are followed to their current chapter automatically.
| Parameter | Type | Description |
|---|---|---|
| slug | string |
{
"name": "get_doc",
"title": "Read a SWATGenX documentation chapter",
"description": "Return one full SWATGenX documentation chapter as markdown, by chapter slug;\n 'overview' is the index and lists every chapter. Retired slugs are followed to their\n current chapter automatically.",
"inputSchema": {
"type": "object",
"title": "get_docArguments",
"properties": {
"slug": {
"type": "string",
"title": "Slug",
"default": "overview"
}
}
},
"annotations": {
"title": "Read a SWATGenX documentation chapter",
"readOnlyHint": true,
"openWorldHint": false,
"idempotentHint": true,
"destructiveHint": false
}
}Where the SWAT+ engine SWATGenX runs comes from (including its parallel/OpenMP build), and how to build it from source. Returns the public fork, the build guide with the exact toolchain, and whether building is needed at all (usually it is not — downloaded models run as-is). No authentication required.
No parameters.
{
"name": "get_engine_info",
"title": "Get SWAT+ engine source, build guide and toolchain",
"description": "Where the SWAT+ engine SWATGenX runs comes from (including its parallel/OpenMP build),\n and how to build it from source. Returns the public fork, the build guide with the exact\n toolchain, and whether building is needed at all (usually it is not — downloaded models run\n as-is). No authentication required.",
"inputSchema": {
"type": "object",
"title": "get_engine_infoArguments",
"properties": {}
},
"annotations": {
"title": "Get SWAT+ engine source, build guide and toolchain",
"readOnlyHint": true,
"openWorldHint": false,
"idempotentHint": true,
"destructiveHint": false
}
}Describe SWATGenX access: accounts by invitation (for a named research collaboration or a funded study), studies quoted for each watershed, and calibration on account credit; how to request access. With an API key, also report the caller's current access and its status.
| Parameter | Type | Description |
|---|---|---|
| api_key | any |
{
"name": "get_access_info",
"title": "Get access level and allocations",
"description": "Describe SWATGenX access: accounts by invitation (for a named research collaboration or a\n funded study), studies quoted for each watershed, and calibration on account credit; how to\n request access. With an API key, also report the caller's current access and its status.",
"inputSchema": {
"type": "object",
"title": "get_access_infoArguments",
"properties": {
"api_key": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"title": "Api Key",
"default": null
}
}
},
"annotations": {
"title": "Get access level and allocations",
"readOnlyHint": true,
"openWorldHint": false,
"idempotentHint": true,
"destructiveHint": false
}
}Start a real SWAT+ model build (authenticated; consumes the caller's fair-use quota). Takes either a USGS gauge id (usgs_station) or a 12-digit HUC12 outlet (huc12_outlet). Returns an order_id for checking the build's status. Typical build: 20 min - 2 h. The API key comes from the X-SWATGenX-Api-Key header on the MCP connection or the api_key argument.
| Parameter | Type | Description |
|---|---|---|
| api_key | any | |
| huc12_outlet | any | |
| usgs_station | any | |
| force_rebuild | boolean |
{
"name": "order_model",
"title": "Build a SWAT+ watershed model",
"description": "Start a real SWAT+ model build (authenticated; consumes the caller's fair-use quota).\n Takes either a USGS gauge id (usgs_station) or a 12-digit HUC12 outlet (huc12_outlet).\n Returns an order_id for checking the build's status. Typical build: 20 min - 2 h. The API key\n comes from the X-SWATGenX-Api-Key header on the MCP connection or the api_key argument.",
"inputSchema": {
"type": "object",
"title": "order_modelArguments",
"properties": {
"api_key": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"title": "Api Key",
"default": null
},
"huc12_outlet": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"title": "Huc12 Outlet",
"default": null
},
"usgs_station": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"title": "Usgs Station",
"default": null
},
"force_rebuild": {
"type": "boolean",
"title": "Force Rebuild",
"default": false
}
}
},
"annotations": {
"title": "Build a SWAT+ watershed model",
"readOnlyHint": false,
"openWorldHint": false,
"idempotentHint": false,
"destructiveHint": false
}
}Check a model build (authenticated): state (queued/running/complete/failed), stage, and timing. order_id is the id order_model returned for the build.
| Parameter | Type | Description |
|---|---|---|
| api_key | any | |
| order_id * | string |
{
"name": "get_order_status",
"title": "Get model build status",
"description": "Check a model build (authenticated): state (queued/running/complete/failed), stage,\n and timing. order_id is the id order_model returned for the build.",
"inputSchema": {
"type": "object",
"title": "get_order_statusArguments",
"required": [
"order_id"
],
"properties": {
"api_key": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"title": "Api Key",
"default": null
},
"order_id": {
"type": "string",
"title": "Order Id"
}
}
},
"annotations": {
"title": "Get model build status",
"readOnlyHint": true,
"openWorldHint": false,
"idempotentHint": true,
"destructiveHint": false
}
}List the caller's models (authenticated): `models`, the model workspaces the account holds (the dashboard's list: site_no, level, vpuid, status flags), and `orders`, the builds requested through the build queue, newest first. `limit` caps each list (default 10, max 50; unbounded listing used to blow the MCP result budget at ~50 builds). `active_only=True` returns only QUEUED/DISPATCHING/RUNNING builds. Each record carries first-class site_no/level/vpuid.
| Parameter | Type | Description |
|---|---|---|
| limit | integer | |
| api_key | any | |
| active_only | boolean |
{
"name": "list_my_models",
"title": "List your models and builds",
"description": "List the caller's models (authenticated): `models`, the model workspaces the account holds\n (the dashboard's list: site_no, level, vpuid, status flags), and `orders`, the builds\n requested through the build queue, newest first. `limit` caps each list (default 10, max 50;\n unbounded listing used to blow the MCP result budget at ~50 builds). `active_only=True`\n returns only QUEUED/DISPATCHING/RUNNING builds. Each record carries first-class\n site_no/level/vpuid.",
"inputSchema": {
"type": "object",
"title": "list_my_modelsArguments",
"properties": {
"limit": {
"type": "integer",
"title": "Limit",
"default": 10
},
"api_key": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"title": "Api Key",
"default": null
},
"active_only": {
"type": "boolean",
"title": "Active Only",
"default": false
}
}
},
"annotations": {
"title": "List your models and builds",
"readOnlyHint": true,
"openWorldHint": false,
"idempotentHint": true,
"destructiveHint": false
}
}Get a download link for a model the caller owns (authenticated). The returned download_url is a TOKEN link valid for 24 hours; fetching it redirects to a storage URL minted fresh per request with a much shorter life (~30 min), so the redirect target expires long before the download_url does. vpuid/level are optional: the server resolves them from site_no (build records also carry them first-class).
| Parameter | Type | Description |
|---|---|---|
| level | any | |
| vpuid | any | |
| api_key | any | |
| site_no * | string |
{
"name": "download_model",
"title": "Get a model download link",
"description": "Get a download link for a model the caller owns (authenticated). The returned\n download_url is a TOKEN link valid for 24 hours; fetching it redirects to a\n storage URL minted fresh per request with a much shorter life (~30 min), so the\n redirect target expires long before the download_url does.\n vpuid/level are optional: the server resolves them from site_no (build records also\n carry them first-class).",
"inputSchema": {
"type": "object",
"title": "download_modelArguments",
"required": [
"site_no"
],
"properties": {
"level": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"title": "Level",
"default": null
},
"vpuid": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"title": "Vpuid",
"default": null
},
"api_key": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"title": "Api Key",
"default": null
},
"site_no": {
"type": "string",
"title": "Site No"
}
}
},
"annotations": {
"title": "Get a model download link",
"readOnlyHint": false,
"openWorldHint": false,
"idempotentHint": true,
"destructiveHint": false
}
}SWATGenX builds ready-to-run SWAT+ hydrological models for gauged and ungauged catchments in the conterminous US, from national datasets. Agencies: studies quoted for each watershed, https://www.swatgenx.com/work-with-us. Accounts are by invitation, for a named research collaboration or a funded study: https://www.swatgenx.com/signup. PFAS monitoring data can be packaged with a model, and PFAS transport modelling is available as a service. This server exposes the same platform that serves https://www.swatgenx.com. WHAT THIS SERVER PROVIDES - Explore existing models: search_swat_models, get_model_calibration (cal/val NSE, PBIAS, method). - Query national data: query_groundwater (wells + lithology). query_pfas serves the PFAS monitoring stations whose source licence allows serving them (the PFAS data API's set), which is NOT national: with `state` the results are US-only; without it every served scope is searched, which may include European stations. Every reply states which set it searched in `scope` and `source`, so a count can be told apart from a US figure. Each answer counts toward the account's monthly allowance of distinct stations and 200 new stations a day; `metering` in the reply says where it stands. - Scope a new model: request_model previews whether a catchment is buildable, read-only. - With the caller's API key: order_model, get_order_status, list_my_models, download_model. - Engine + build details: get_engine_info. Access: get_access_info. DOCUMENTATION - search_docs(query) then get_doc(slug) serve the full SWATGenX I/O reference: model bundle layout, SWAT+ input files, PFAS inputs, engine run modes and NetCDF output, MODFLOW-6 coupling, calibration artifacts, and service limits. Same text as https://www.swatgenx.com/docs. The schemas are transcribed from the engine readers/writers, and stock SWAT+ documentation does not cover the extensions. PYTHON PACKAGE - `pip install swatgenx` gives the same platform as a library, for users who want to work in a notebook or script rather than through an agent: import swatgenx as sg sg.catalog(state="MI") # Michigan's example models: site_no, vpuid, n_channels c = sg.Client(api_key="..."); c.order(usgs_station="04124500") It serves reproducible or scripted access rather than a one-off answer. CONVENTIONS - USGS site numbers keep their leading zero ("01010000"); the server accepts either spelling. - The model, data and build tools need model access; accounts are by invitation. Without it: the documentation, access and engine-info tools. - Build requests and downloads additionally need the caller's API key. - This server is read-only apart from the explicitly authenticated build tools.
Names, descriptions and schemas were written by the server and are shown as data.
Pattern matches on what the server declares. They show what it says it can do, not what it does.
None yet. Changes appear after a server's second daily check.