Metadata-Version: 2.4
Name: ph-civic-data-mcp
Version: 0.4.0
Summary: MCP server for Philippine civic data: PSGC, infra spending accountability, PSA economy (inflation, labor, poverty, population), one-call area profiles, earthquakes, weather, typhoons, procurement, solar/climate, air quality, vegetation
Project-URL: Homepage, https://github.com/xmpuspus/ph-civic-data-mcp
Project-URL: Repository, https://github.com/xmpuspus/ph-civic-data-mcp
Project-URL: Issues, https://github.com/xmpuspus/ph-civic-data-mcp/issues
Author: Xavier Puspus
License: MIT
License-File: LICENSE
Keywords: accountability,civic-tech,ibtracs,mcp,modis,nasa,pagasa,philgeps,philippines,phivolcs,psa,psgc,usgs,world-bank
Classifier: Development Status :: 4 - Beta
Classifier: Intended Audience :: Developers
Classifier: License :: OSI Approved :: MIT License
Classifier: Operating System :: OS Independent
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Classifier: Topic :: Scientific/Engineering :: Information Analysis
Requires-Python: >=3.11
Requires-Dist: beautifulsoup4>=4.12.0
Requires-Dist: cachetools>=5.3.0
Requires-Dist: fastmcp<4.0.0,>=3.0.0
Requires-Dist: httpx>=0.27.0
Requires-Dist: lxml>=5.0.0
Requires-Dist: openpyxl>=3.1.0
Requires-Dist: pydantic>=2.0.0
Requires-Dist: python-dateutil>=2.9.0
Provides-Extra: dev
Requires-Dist: pytest-asyncio>=0.23.0; extra == 'dev'
Requires-Dist: pytest>=8.0.0; extra == 'dev'
Requires-Dist: ruff>=0.6.0; extra == 'dev'
Description-Content-Type: text/markdown

# ph-civic-data-mcp

<!-- mcp-name: io.github.xmpuspus/ph-civic-data-mcp -->

> The multi-source MCP server for Philippine civic data. PSGC codes, infra spending accountability, earthquakes, weather, typhoons, procurement, population, poverty, solar radiation, air quality, satellite vegetation indices, and macro indicators — all in your AI agent, no API keys required.

[![PyPI](https://img.shields.io/pypi/v/ph-civic-data-mcp.svg)](https://pypi.org/project/ph-civic-data-mcp/)
[![Python 3.11+](https://img.shields.io/badge/python-3.11+-blue.svg)](https://www.python.org/downloads/)
[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
[![Glama AAA](https://glama.ai/mcp/servers/xmpuspus/ph-civic-data-mcp/badges/score.svg)](https://glama.ai/mcp/servers/xmpuspus/ph-civic-data-mcp)
[![MCP Registry](https://img.shields.io/badge/MCP%20Registry-io.github.xmpuspus%2Fph--civic--data--mcp-blue)](https://registry.modelcontextprotocol.io/v0.1/servers?search=ph-civic-data-mcp)

**One-click install:**
[![Add to Cursor](https://img.shields.io/badge/Add%20to-Cursor-000000?logo=cursor)](cursor://anysphere.cursor-deeplink/mcp/install?name=ph-civic-data&config=eyJjb21tYW5kIjogInV2eCIsICJhcmdzIjogWyJwaC1jaXZpYy1kYXRhLW1jcCJdfQ==)
[![Add to VS Code](https://img.shields.io/badge/Add%20to-VS%20Code-007ACC?logo=visualstudiocode)](https://insiders.vscode.dev/redirect/mcp/install?name=ph-civic-data&config=%7B%22command%22%3A%22uvx%22%2C%22args%22%3A%5B%22ph-civic-data-mcp%22%5D%7D)
[![Install via Smithery](https://img.shields.io/badge/Install%20via-Smithery-blueviolet)](https://smithery.ai/server/ph-civic-data-mcp)
[![Add via Claude Code](https://img.shields.io/badge/Add%20via-Claude%20Code-D97757?logo=anthropic)](https://code.claude.com/docs/en/mcp)

`ph-civic-data-mcp` is a zero-cost, `stdio`-transport MCP server. **v0.4.0** expands the PSA OpenSTAT layer beyond population/poverty into the live economy — `get_inflation_stats` (regional CPI), `get_labor_stats` (Labor Force Survey), `get_health_indicators` — and adds the auto-stitch differentiator `get_area_profile`: name a place once and get the resolved PSGC spine plus demographics, economy, procurement, hazard, and weather composed in a single agent turn, with infrastructure notices already normalized per 100k residents. **v0.3.1** was a correctness pass on the v0.3.0 accountability layer (no fabricated PAGASA advisories, stoplisted hazard tokens, "City of Manila" coordinate bridge, province/agency-alias infra search). **v0.3.0** added the PH Accountability layer: PSGC location resolver, infra spending search, and one cross-source heuristic that flags procurement notices for further review by cross-referencing PHIVOLCS earthquakes and PAGASA typhoon footprints. **v0.2.0** added six no-auth scientific + open-data sources (NASA POWER, Open-Meteo Air Quality, NASA MODIS, USGS FDSN, NOAA IBTrACS, World Bank Open Data) on top of the original four Philippine government feeds (PHIVOLCS, PAGASA, PhilGEPS, PSA). 29 tools total. Boots and runs with zero API keys.

All data sourced from public records (PSGC, PHIVOLCS, PAGASA, PhilGEPS, PSA, and open scientific feeds). Heuristic indicators are statistical only; specific allegations, if any, require independent investigation and corroboration.

## This is how easy it is to set up

One JSON file. One `claude` command. Your agent just correlated live Philippine weather with 2020 Census population data in a single turn.

![setup](docs/demo_setup.gif)

The recording above isn't scripted. It's `vhs docs/demo_setup.tape`, which spawns Claude Code with `--mcp-config` pointing at this server, and Claude fans out in parallel to `get_weather_forecast` (Open-Meteo) and `get_population_stats` (PSA PXWeb), then correlates them. The temperatures (30.4 / 30.9 / 31.0 °C max over Apr 19-21) and NCR population (13,484,462) in the streamed answer are what the live sources returned at the moment of the recording.

Works the same way in Claude Desktop, Cursor, Zed, VS Code, or any MCP-compatible client. One `"command": "uvx"`, one `"args": ["ph-civic-data-mcp"]`, done.

## Demo

Every GIF below is a real VHS recording of `docs/live_demo.py`. It spawns `uvx ph-civic-data-mcp` from this PyPI release and calls each tool over the real MCP stdio protocol. The panels you see contain the actual JSON returned by the server. Nothing is staged.

A grand tour hitting 7 tools across all 4 sources in one session:

![grand tour](docs/demo.gif)

Per-source walkthroughs below. To reproduce any of them locally: `uv run python docs/live_demo_single.py <suite>`.

## Why this exists

Philippine civic-data portals publish open data, but each in its own schema — scraped HTML tables, PXWeb JSON, undocumented APIs. Nothing ties them together for an AI agent. This server does.

A handful of other Philippine civic-data MCP servers exist (PSGC administrative geography, holidays, DHSUD license-to-sell, DepEd schools), each covering one dataset. None expose hazard feeds, weather, procurement, or statistical data, and none combine sources. This server does both. See the Prior art section below for the full list.

## Install

```bash
uvx ph-civic-data-mcp
```

Or via pip:

```bash
pip install ph-civic-data-mcp
```

## Setup

### Claude Desktop

Add to `~/Library/Application Support/Claude/claude_desktop_config.json`:

```json
{
  "mcpServers": {
    "ph-civic-data": {
      "command": "uvx",
      "args": ["ph-civic-data-mcp"]
    }
  }
}
```

### Claude Code

Add to `.claude/settings.json`:

```json
{
  "mcpServers": {
    "ph-civic-data": {
      "command": "uvx",
      "args": ["ph-civic-data-mcp"]
    }
  }
}
```

Or install via the Claude Code CLI:

```bash
claude mcp add ph-civic-data -- uvx ph-civic-data-mcp
```

### Cursor, Zed, other MCP clients

Any client that supports the stdio MCP transport works. Point the command at `uvx ph-civic-data-mcp`. No API keys required for the default configuration.

## What you can ask

After setup, ask your agent:

- _"How hot is Metro Manila this week and how many people are affected?"_
- _"What earthquakes happened in the Philippines in the last 24 hours?"_
- _"Is Taal volcano active right now?"_
- _"What's the 3-day weather forecast for Quezon City?"_
- _"Are there active typhoons in the Philippines right now?"_
- _"Search PhilGEPS for flood control contracts."_
- _"What is the population of Region VII based on the PSA?"_
- _"What is the poverty incidence in the Bicol Region?"_
- _"Give me a multi-hazard risk profile for Leyte."_
- _"What's the solar irradiance in Ilocos Norte this week? Good site for a PV farm?"_ **(v0.2.0)**
- _"Compare air quality in Makati and Cebu City right now."_ **(v0.2.0)**
- _"What do MODIS NDVI composites say about vegetation health over the Nueva Ecija rice bowl?"_ **(v0.2.0)**
- _"Cross-check the magnitudes that PHIVOLCS and USGS assigned to last week's events."_ **(v0.2.0)**
- _"List all typhoons that passed through the PAR in the 2024 season."_ **(v0.2.0)**
- _"What's the Philippines' GDP growth and poverty ratio over the last decade?"_ **(v0.2.0)**
- _"Resolve 'Sta. Mesa, Manila' to its official PSGC code."_ **(v0.3.0)**
- _"Find DPWH flood-control or construction projects in Pampanga over P50M with low progress."_ **(v0.3.0)**
- _"Show me PH infrastructure spending breakdown by category for the latest PhilGEPS window."_ **(v0.3.0)**
- _"Are there any infra projects whose locations overlap a recent earthquake or typhoon footprint? Flag them for review."_ **(v0.3.0)**
- _"Walk the location hierarchy from PSGC code 072217000 up to its region."_ **(v0.3.0)**
- _"What's the cache TTL and freshness of every data source this server uses?"_ **(v0.3.0)**
- _"What's the latest inflation rate in Central Visayas vs nationally?"_ **(v0.4.0)**
- _"What are the current employment and underemployment rates from the PSA Labor Force Survey?"_ **(v0.4.0)**
- _"What's the maternal mortality ratio and total fertility rate for the Philippines?"_ **(v0.4.0)**
- _"Give me the full civic profile of Cebu — population, poverty, inflation, jobs, procurement activity, hazards, and weather, in one shot."_ **(v0.4.0)**

## Per-source demos

### PHIVOLCS — earthquakes + volcano alert levels
![phivolcs](docs/demo_phivolcs.gif)

### PAGASA — weather forecast + typhoon tracking
![pagasa](docs/demo_pagasa.gif)

### PhilGEPS — procurement search + aggregation
![philgeps](docs/demo_philgeps.gif)

### PSA — population (2020 Census) + poverty (2023 Full-Year)
![psa](docs/demo_psa.gif)

### Cross-source — parallel multi-hazard risk profile
![cross-source](docs/demo_combined.gif)

### PSA economy + auto-stitch — inflation, labor, health, one-call area profile *(v0.4.0)*
![v0.4.0 per-source](docs/demo_v040.gif)

The LinkedIn showcase — one question correlated across five sources through
`get_area_profile` — is under [What's new in v0.4.0](#whats-new-in-v040--psa-economy--the-auto-stitch-context-layer).

### How the demos are produced

`docs/live_demo.py` and `docs/live_demo_single.py` open an MCP `StdioTransport` pointing at `uvx ph-civic-data-mcp` (which resolves to this PyPI release), call the tools, and render the responses with [Rich](https://github.com/Textualize/rich) (panels, tables, syntax-highlighted JSON, live spinners). [`vhs`](https://github.com/charmbracelet/vhs) drives a real terminal and records the session. Tapes are committed under `docs/*.tape`.

## Data sources

| Source | Data | Update frequency | Auth |
|---|---|---|---|
| PHIVOLCS | Earthquakes, bulletins, volcano alerts | 5 min (earthquakes), 30 min (volcanoes) | None |
| PAGASA | 10-day weather, active typhoons, alerts | Hourly | Optional `PAGASA_API_TOKEN` |
| Open-Meteo | Weather fallback when PAGASA token absent | Hourly | None |
| PhilGEPS | Government procurement notices (latest ~100) | 6 h (cached) | None |
| PSA OpenSTAT | Population (2020 Census), poverty (2023), CPI/inflation, Labor Force Survey, health indicators | Per-table vintage | None |
| **NASA POWER** *(v0.2.0)* | Daily solar irradiance + temp/precip/wind, any lat/lng | Daily | None |
| **Open-Meteo Air Quality** *(v0.2.0)* | PM2.5/PM10/NO2/SO2/O3/CO + AQI | Hourly | None |
| **NASA MODIS via ORNL DAAC** *(v0.2.0)* | NDVI/EVI vegetation indices (250m, 16-day composites) | Weekly | None |
| **USGS FDSN** *(v0.2.0)* | Philippine-region earthquakes from global seismic network | Minutes | None |
| **NOAA IBTrACS** *(v0.2.0)* | Historical tropical cyclone tracks through the PAR | Per storm | None |
| **World Bank Open Data** *(v0.2.0)* | Philippine macro indicators (GDP, poverty ratio, inflation, etc.) | Annual | None |
| **PSGC** *(v0.3.0)* | Philippine Standard Geographic Code via [psgc.gitlab.io](https://psgc.gitlab.io/api/) (PSA dataset mirror) | When PSA publishes a new version | None |
| **PH Infra (PhilGEPS-backed)** *(v0.3.0)* | Filtered infra notices for construction / road / bridge / flood control | 6 h cache window | None |
| **PSA economy** *(v0.4.0)* | CPI/inflation (regional), Labor Force Survey rates, health indicators via PXWeb browse-discovery | Per-table vintage (read from each table) | None |
| **Area profile (auto-stitch)** *(v0.4.0)* | One-call composition: PSGC + PSA + PhilGEPS + PHIVOLCS + PAGASA, with per-capita normalization | Live per request; 1 h cache | None |

## All tools

| Tool | Description | Key params |
|---|---|---|
| `get_latest_earthquakes` | Recent PH earthquakes | `min_magnitude`, `limit`, `region` |
| `get_earthquake_bulletin` | Full PHIVOLCS bulletin for one event | `bulletin_url` |
| `get_volcano_status` | Alert level per monitored PH volcano | `volcano_name` |
| `get_weather_forecast` | 1–10 day forecast (PAGASA or Open-Meteo) | `location`, `days` |
| `get_active_typhoons` | Active tropical cyclones in/near PAR | — |
| `get_weather_alerts` | Active PAGASA warnings | `region` |
| `search_procurement` | Keyword search on PhilGEPS notices | `keyword`, `agency`, `region`, `date_from/to`, `limit` |
| `get_procurement_summary` | Aggregate procurement stats | `agency`, `region`, `year` |
| `get_population_stats` | 2020 Census population | `region` |
| `get_poverty_stats` | 2023 Full-Year poverty incidence | `region` |
| `assess_area_risk` | Multi-hazard profile (parallel PHIVOLCS + PAGASA) | `location` |
| **`get_solar_and_climate`** *(v0.2.0)* | NASA POWER daily solar irradiance + climate variables at any coordinate | `latitude`, `longitude`, `start_date`, `end_date` |
| **`get_air_quality`** *(v0.2.0)* | Real-time air quality for ~80 major PH cities via Open-Meteo | `location` |
| **`get_vegetation_index`** *(v0.2.0)* | MODIS NDVI + EVI vegetation index timeseries at any coordinate | `latitude`, `longitude`, `start_date`, `end_date` |
| **`get_usgs_earthquakes_ph`** *(v0.2.0)* | PH-bbox earthquakes from USGS global network (cross-ref to PHIVOLCS) | `start_date`, `end_date`, `min_magnitude`, `limit` |
| **`get_historical_typhoons_ph`** *(v0.2.0)* | Historical typhoons that passed through the Philippine AOR (IBTrACS) | `year`, `limit` |
| **`get_world_bank_indicator`** *(v0.2.0)* | Philippine macro indicator from World Bank Open Data (code or friendly alias) | `indicator`, `per_page` |
| **`resolve_ph_location`** *(v0.3.0)* | Fuzzy-resolve a free-text PH place name to its canonical PSGC record | `query` |
| **`list_admin_units`** *(v0.3.0)* | Browse children of a PSGC node, or top-level regions when `parent_code` is None | `parent_code`, `level`, `limit` |
| **`get_location_hierarchy`** *(v0.3.0)* | Full chain region -> province -> city/municipality for one PSGC code | `psgc_code` |
| **`search_infra_projects`** *(v0.3.0)* | Filter PhilGEPS notices for infra-related work (construction, road, bridge, flood control) | `keyword`, `region`, `province`, `year`, `min_cost_php`, `status`, `limit` |
| **`get_infra_project`** *(v0.3.0)* | Full record for one infra project by `project_id` | `project_id` |
| **`summarize_infra_spending`** *(v0.3.0)* | Aggregate infra notice stats by category, region, agency | `region`, `year`, `funding_source` |
| **`flag_infra_anomalies`** *(v0.3.0)* | Heuristic indicators for further review (high_cost_no_progress, hazard_overlap, duplicate_titles_same_agency) | `region`, `province`, `min_cost_php` |
| **`get_data_freshness`** *(v0.3.0)* | Catalog of every upstream source with TTL, freshness, license | (none) |
| **`get_inflation_stats`** *(v0.4.0)* | Headline year-on-year CPI inflation, national or regional, latest published month | `area` |
| **`get_labor_stats`** *(v0.4.0)* | PSA Labor Force Survey key rates (LFPR, employment, unemployment, underemployment) | `region` |
| **`get_health_indicators`** *(v0.4.0)* | National health indicators (maternal mortality, total fertility rate, browse-discovered set) | `indicator` |
| **`get_area_profile`** *(v0.4.0)* | One-call auto-stitch: resolved PSGC + demographics + economy + procurement + hazard + weather, per-capita normalized | `location` |

## Environment variables

| Variable | Required | Notes |
|---|---|---|
| `PAGASA_API_TOKEN` | Optional | Requires formal PAGASA request. Without it, weather auto-falls-back to Open-Meteo. |

No mandatory API keys. The server boots and all 25 tools work without any token.

## Data freshness warnings

- **Population:** 2020 Census. No later national data exists yet.
- **Poverty:** 2023 Full-Year poverty statistics (latest PSA release).
- **Procurement:** PhilGEPS open data does not expose filterable search externally. This server scrapes the latest ~100 bid notices and filters client-side. Cached 6h.
- **Emergencies:** for real-time disaster response, always check [ndrrmc.gov.ph](https://ndrrmc.gov.ph) and official PHIVOLCS/PAGASA channels. This server is for research, not life-safety decisions.

## Architecture

- Python 3.11+, `fastmcp>=3.0.0,<4.0.0`
- Two HTTP clients: standard + `PHIVOLCS_CLIENT` with `verify=False` (PHIVOLCS has a broken SSL cert chain). SSL verification is **never** disabled globally.
- In-memory TTL caches per source; no disk writes.
- stdio transport only (zero hosting cost).
- PSA table paths are discovered via the PXWeb browse API, never hardcoded.

## Development

```bash
git clone https://github.com/xmpuspus/ph-civic-data-mcp
cd ph-civic-data-mcp
uv sync --extra dev

# MCP Inspector
fastmcp dev src/ph_civic_data_mcp/server.py

# Tests (run against live APIs)
uv run pytest tests/ -v

# Build
uv run python -m build
uv run twine check dist/*
```

## Limitations

- **PAGASA token is gated.** Non-government users may be denied. Open-Meteo fallback removes this as a hard dependency.
- **PhilGEPS is not real-time.** Public portal exposes no filterable API; this server operates on the latest ~100 notices with client-side filtering.
- **Emergencies:** direct users to official channels; this is a research tool.

## What's new in v0.4.0 — PSA economy + the auto-stitch context layer

v0.4.0 does two things: it takes the PSA OpenSTAT layer past population/poverty
into the live economy, and it adds the differentiator the project was building
toward — a single tool that hands the agent correlated multi-source context in
one turn instead of making it orchestrate eight calls.

### New PSA economy tools

All three use the same browse-discovery convention as the existing
population/poverty tools: only the stable subject path is fixed, the `.px`
table is discovered by text (never a hardcoded id), and the data vintage is
read from each table's own time dimension — never from the response
timestamp, which is just server wall-clock.

- **`get_inflation_stats(area)`** — headline year-on-year CPI inflation, 2018
  base, national or by region. PSA splits long series into era tables with
  near-identical titles (a backcasted 1958–1994 table sits right next to the
  current one); the resolver picks the table whose time dimension reaches the
  most recent year, so you always get the current series. Reports the exact
  reference period because PSA publishes monthly with a lag.
- **`get_labor_stats()`** — Labor Force Survey key rates: labor-force
  participation, employment, unemployment, underemployment. National (the PSA
  key-indicator table has no regional split; a `region` argument is recorded
  as an explicit caveat rather than silently ignored).
- **`get_health_indicators(indicator)`** — national health indicators
  (maternal mortality ratio, total fertility rate) with the full available set
  browse-discovered, not hardcoded.

### The auto-stitch layer: `get_area_profile(location)`

Name a place once. The tool resolves it to its PSGC code, then fans out in
parallel and returns demographics, economy, procurement activity, multi-hazard
risk, and the short-range weather outlook in one envelope — and it does the
cross-source normalization the agent would otherwise have to do itself
(infrastructure notices per 100k residents, each block carrying its own
reference period). One round-trip replaces about eight, and the agent never
has to know that population keys on region while procurement keys on province
and hazard is PHIVOLCS+PAGASA.

### The showcase: one question, five sources, a defensible read

![v0.4.0 hero](docs/demo_v040_hero.gif)

This is a real `claude -p --mcp-config` turn (tape: `docs/demo_v040_hero.tape`).
One question:

> _"Which region's recent government infra-notice count per 100k looks least
> proportionate to its economic need — weigh poverty, regional inflation, and
> population — and note hazard exposure. Flagged-for-review language only."_

The agent calls `get_area_profile` twice — Eastern Visayas and Central Visayas
— each composing PSGC + PSA + PhilGEPS + PHIVOLCS + PAGASA, then reasons across
the two. Values are what the live sources returned at capture (2026-05-18):

- **Eastern Visayas**: poverty 20.3% (2023), population 4.55M, regional
  inflation 8.5% (2026 Apr), earthquake risk Low, 0 infra notices in the
  current PhilGEPS window.
- **Central Visayas**: poverty 12.3% (2023), population 8.08M, regional
  inflation 10.8% (2026 Apr), earthquake risk Low, 0 infra notices.
- **The read it produced**: Eastern Visayas _warrants closer review as the more
  underserved of the two_ — higher poverty against a smaller population, yet the
  same zero-notice procurement footprint as the wealthier, faster-inflating
  Central Visayas; comparable (low) hazard exposure, so the flat footprint is
  not explained by a disaster driver. "An absence that warrants further
  investigation … not a finding of wrongdoing", with `source_url` and the
  public-data disclaimer attached.

That conclusion is impossible from any single feed: it needs census population
+ PSA poverty + PSA regional inflation (the v0.4.0 economic denominator) +
PhilGEPS procurement + PHIVOLCS/PAGASA hazard, normalized per capita. The new
auto-stitch layer makes the agent's orchestration invisible — one question, one
defensible answer.

### Per-source demo

![v0.4.0 per-source](docs/demo_v040.gif)

`docs/live_demo_v040.py` over the real MCP stdio protocol (tape:
`docs/demo_v040_sources.tape`): national + regional headline inflation, Labor
Force Survey rates, national health indicators, and the one-call
`get_area_profile`. Every panel is live tool JSON from the server.

### Tests

`tests/test_psa_expansion.py` (8) and `tests/test_autostitch.py` (4) — live
integration tests mirroring `tests/test_phivolcs.py`, covering real-figure
sanity bands, the national-only labor caveat, graceful unknown-area/unknown
-indicator paths, the unresolved-location degradation path, and per-capita
arithmetic consistency. The stale v0.3.0 `server_version` assertion in
`tests/test_v030_live.py` was repinned to the package `__version__`.

## What's new in v0.3.1 — correctness pass on the v0.3.0 accountability layer

### Re-recorded demo: "Sta. Mesa, Manila" + "flood control in Pampanga"

![accountability demo](docs/demo_accountability.gif)

One real `claude -p --mcp-config` call against `uvx ph-civic-data-mcp@0.3.1` (live PyPI), exercising the v0.3.1 fixes in a single turn:

1. `resolve_ph_location("Sta. Mesa, Manila")` resolves cleanly to PSGC "City of Manila" — the v0.3.0 chain silently failed here because `city_to_coords` couldn't invert "City of Manila" back to a coordinate. v0.3.1's `city_to_coords` strips the `city of ` / `municipality of ` prefixes and walks comma-segments.
2. `search_infra_projects(keyword="flood control", province="Pampanga")` returns matches via the new `_PROVINCE_AGENCY_HINTS` map, which expands "Pampanga" to also catch DPWH agency names like "REGION III" / "Central Luzon" / "San Fernando". v0.3.0 returned a false-empty list because the substring filter only checked title + agency, never the regional aliases. The hint map covers all 81 PH provinces plus NCR.
3. `flag_infra_anomalies(province="Pampanga")` now emits clean `hazard_overlap` results because the new `_proper_noun_tokens` helper requires capitalisation in the source string and applies an explicit stoplist of geographic chrome (`city`, `region`, `eastern`, `philippines`, ...). v0.3.0 fired on tokens like `['city']` because every alpha word ≥4 chars from earthquake locations went into the keyword set — the README itself had to caveat this in the v0.3.0 demo caption. The apologetic caveat is gone.

Tape: [`docs/demo_accountability.tape`](docs/demo_accountability.tape) (35s VHS recording, mpdecimate post-process). The gif is the actual frames vhs captured against the live PyPI release.

### What changed under the hood

- **`get_weather_alerts` no longer fabricates advisories.** v0.3.0's regex matched alert names ("Heavy Rainfall Warning", "Flood Advisory", "Gale Warning") wherever they appeared on the PAGASA homepage including the navigation menu and breadcrumbs. v0.3.1 returns `[]` when the page is reachable but the active-warning state is ambiguous, and `[]` with the explicit "No Active Warnings" signal when the homepage says so. For real-time advisories, hit `bagong.pagasa.dost.gov.ph` directly.
- **`get_volcano_status` not-found branch** emits `source_url` + `license` + `caveats` for envelope parity with the rest of v0.3.0.
- **Server `instructions` block** now anchors civic-tech framing every turn: agents are instructed to use defensible language ("flagged for review"), never accusations, and to cite `source_url` for every factual claim.
- **`get_data_freshness` doubles as health/version probe.** Response now includes `server_name`, `transport`, and `tool_count`.
- **urllib3 `InsecureRequestWarning`** suppressed for the dedicated PHIVOLCS client only. Verify-disabled scope is unchanged.

### Distribution

- **One-click install badges** in the header (Cursor, VS Code, Smithery, Claude Code).
- **Claude Desktop `.mcpb` bundle** attached to the [v0.3.1 GitHub release](https://github.com/xmpuspus/ph-civic-data-mcp/releases/tag/v0.3.1) (2.8 MB). Double-click to install. Optional PAGASA token prompted via `user_config`.
- **`.github/workflows/ci.yml`** runs ruff + tests on every push and PR (Python 3.11 and 3.12).
- **`.github/workflows/release-smoke.yml`** fires on every `v*.*.*` tag, installs the freshly-published wheel from PyPI in a fresh venv, and asserts that `len(tools) >= 25` plus offline regression checks for the v0.3.1 geo fixes.
- **`tests/test_v031_fixes.py`** — 13 new regression tests pinning each fix. 70 tests pass total (57 existing + 13 new).

### DPWH portal status (verified 2026-05-02)

The DPWH transparency portal at `transparency.dpwh.gov.ph` and `api.transparency.dpwh.gov.ph` is still behind a Cloudflare bot challenge (HTTP 403, "Just a moment..."). PhilGEPS remains the source of record for infra-spending tools; the single integration point in `sources/infra.py` is unchanged and ready to swap in when DPWH lifts the block.

## What's new in v0.3.0 — PH Accountability layer

This release adds three tightly-scoped capabilities for civic accountability work, plus one polish tool.

1. **PSGC backbone** (`resolve_ph_location` / `list_admin_units` / `get_location_hierarchy`) — fuzzy free-text place name resolution to the canonical Philippine Standard Geographic Code, full hierarchy walks, and admin-unit browsing. Sourced from the community-mirrored PSA dataset at [psgc.gitlab.io](https://psgc.gitlab.io/api/).
2. **Infra spending** (`search_infra_projects` / `get_infra_project` / `summarize_infra_spending`) — PhilGEPS notices filtered for construction / road / bridge / flood control / drainage / school building / civil works. The DPWH Transparency portal at `transparency.dpwh.gov.ph` is currently behind Cloudflare's bot challenge and not reachable to non-browser clients, so v0.3.0 sources from the open PhilGEPS listing instead.
3. **Cross-source anomaly indicator** (`flag_infra_anomalies`) — emits heuristic flags for further review by cross-referencing the infra notice window against PHIVOLCS earthquakes (>=M4.0 in last 30d) and active PAGASA typhoon footprints. Three rules: `high_cost_no_progress`, `hazard_overlap`, `duplicate_titles_same_agency`. Every flagged item ships with a "Statistical indicators derived from public data. Patterns may have legitimate explanations." disclaimer.
4. **Polish** — `get_data_freshness` returns the catalog of every upstream source with cache TTL, freshness expectation, and license. Every new tool response includes `source`, `source_url`, `data_retrieved_at`, and `license`.

### Per-tool live outputs

Every JSON block below is the **actual** response from each new tool, captured by running `uv run python docs/live_probe_v030.py` against live public APIs on the release date. Lists are clipped to fit; full output saved to `/tmp/live_probe_v030_output.json`.

#### `resolve_ph_location` — PSGC fuzzy resolver

Handles common patterns: comma-separated qualifiers ("Sta. Mesa, Manila"), Filipino abbreviations (Sta., Sto., Brgy.), partial names ("Pampanga"), full names.

```
$ resolve_ph_location(query="Sta. Mesa, Manila")
```

```json
{
  "psgc_code": "133900000",
  "name": "City of Manila",
  "level": "city",
  "parent_code": "130000000",
  "region_code": "130000000",
  "island_group": "luzon",
  "matched": true,
  "match_score": 0.893,
  "source": "PSGC",
  "source_url": "https://psgc.gitlab.io/api/cities-municipalities/133900000/",
  "license": "Public domain (PSA Philippine Standard Geographic Code)"
}
```

#### `get_location_hierarchy` — full chain region -> province -> city/municipality

```
$ get_location_hierarchy(psgc_code="072200000")
```

```json
{
  "psgc_code": "072200000",
  "chain": [
    {
      "psgc_code": "070000000",
      "name": "Central Visayas",
      "level": "region",
      "source_url": "https://psgc.gitlab.io/api/regions/070000000/"
    },
    {
      "psgc_code": "072200000",
      "name": "Cebu",
      "level": "province",
      "source_url": "https://psgc.gitlab.io/api/provinces/072200000/"
    }
  ],
  "source": "PSGC",
  "license": "Public domain (PSA Philippine Standard Geographic Code)"
}
```

#### `search_infra_projects` — PhilGEPS notices, infra-only

```
$ search_infra_projects(keyword="construction", limit=3)
```

```json
[
  {
    "project_id": "23164",
    "title": "Construction of Sewage Treatment Plant (STP) in Pasig Bliss Village III ...",
    "agency": "CITY OF PASIG",
    "category": "civil works (other)",
    "cost_php": null,
    "currency": "PHP",
    "status": "Open",
    "date_published": "2026-04-27",
    "source": "PhilGEPS",
    "source_url": "https://www.philgeps.gov.ph/",
    "license": "Public — PhilGEPS open notice listing"
  },
  {
    "project_id": "23319",
    "title": "26C00029 Asset Preservation Program ... Reconstruction Upgrading ...",
    "agency": "DEPARTMENT OF PUBLIC WORKS AND HIGHWAYS - REGION I",
    "category": "road / highway",
    "cost_php": null,
    "status": "Open",
    "source_url": "https://www.philgeps.gov.ph/"
  }
]
```

#### `summarize_infra_spending` — aggregated breakdown

```
$ summarize_infra_spending()
```

```json
{
  "total_count": 15,
  "total_value_php": null,
  "by_category": {
    "road / highway": 7,
    "civil works (other)": 6,
    "bridge": 1,
    "school building": 1
  },
  "by_funding_source": {"unknown": 15},
  "reference_period": {"from": "2026-04-27", "to": "2026-04-27"},
  "note": "Computed over the latest infra-keyword-matched PhilGEPS notice window (cached 6h). Approved budget totals are not published in the open notice listing, so total_value_php is typically null.",
  "source": "PhilGEPS",
  "source_url": "https://www.philgeps.gov.ph/",
  "license": "Public — PhilGEPS open notice listing",
  "disclaimer": "Statistical indicators derived from public data. Patterns may have legitimate explanations."
}
```

#### `flag_infra_anomalies` — heuristic indicators across PhilGEPS + PHIVOLCS + PAGASA

```
$ flag_infra_anomalies(min_cost_php=50_000_000)
```

```json
{
  "filters": {"region": null, "province": null, "min_cost_php": 50000000},
  "projects_examined": 15,
  "flagged_count": 4,
  "rules_summary": {"hazard_overlap": 4},
  "flagged": [
    {
      "project_id": "23164",
      "title": "Construction of Sewage Treatment Plant (STP) ... Pasig City",
      "agency": "CITY OF PASIG",
      "rule_fired": "hazard_overlap",
      "evidence": "project title overlaps with recent hazard footprint keywords: ['city']",
      "source_url": "https://www.philgeps.gov.ph/"
    }
  ],
  "hazard_inputs": {"recent_earthquake_count_30d": 0, "active_typhoon_count": 0},
  "source": "PhilGEPS + PHIVOLCS + PAGASA",
  "source_url": "https://www.philgeps.gov.ph/, https://earthquake.phivolcs.dost.gov.ph/, https://bagong.pagasa.dost.gov.ph/",
  "license": "Public — PhilGEPS, PHIVOLCS, PAGASA notice and bulletin pages",
  "disclaimer": "Statistical indicators derived from public data. Patterns may have legitimate explanations."
}
```

Each flag is a heuristic **indicator**, not an accusation. The `hazard_overlap` rule simply says the project title shares keywords with a recent hazard footprint; the project may be entirely legitimate post-disaster reconstruction. Treat output as a starting point for further investigation, not as evidence of wrongdoing.

#### `get_data_freshness` — TTL and license catalog

```
$ get_data_freshness()
```

```json
{
  "server_version": "0.3.0",
  "asof": "2026-04-27T00:09:21+00:00",
  "sources": [
    {"source": "PSGC", "source_url": "https://psgc.gitlab.io/api/", "freshness": "Updated when PSA publishes new PSGC version (annual or quarterly)", "cache_ttl_seconds": 86400, "license": "Public domain (PSA Philippine Standard Geographic Code)"},
    {"source": "PHIVOLCS earthquakes", "source_url": "https://earthquake.phivolcs.dost.gov.ph/", "freshness": "5-minute table refresh; bulletins published per event", "cache_ttl_seconds": 300, "license": "Public — PHIVOLCS public bulletin pages"}
  ]
}
```

## What's new in v0.2.0 — six new no-auth sources

### Correlation demo: "Why is Metro Manila so hot right now?"

![correlation demo](docs/demo_correlation.gif)

One unscripted `claude -p --mcp-config` call, real MCP stdio transport, live upstream APIs. Claude picks three sources out of the 17 tools — **PAGASA/Open-Meteo 7-day forecast** (v0.1.x), **NASA POWER solar irradiance** (v0.2.0, new), and **Open-Meteo Air Quality** (v0.2.0, new) — then correlates them into a three-sentence answer. The numbers in the response (6.8–7.3 kWh/m²/day irradiance, 30.3–31.8°C daytime temps, PM2.5 24.8 µg/m³, US AQI 91) are exactly what the live endpoints returned at the moment of recording. Tape: `docs/demo_correlation.tape`.

### Per-tool live outputs

Every JSON block below is the **actual** response from each tool, captured by running `uv run python docs/live_probe_v020.py` against live public APIs on the release date. No placeholders, no truncation tricks — lists were clipped to fit.

### `get_solar_and_climate` — NASA POWER

Daily solar irradiance + climate at any coordinate. Useful for PV siting and agricultural modeling.

```
$ get_solar_and_climate(latitude=14.5995, longitude=120.9842,
                        start_date="2026-04-01", end_date="2026-04-07")
```

```json
{
  "latitude": 14.5995,
  "longitude": 120.9842,
  "start_date": "2026-04-01",
  "end_date": "2026-04-07",
  "days": [
    {"date": "2026-04-01", "solar_irradiance_kwh_m2": 6.79, "temp_c": 25.4, "precipitation_mm": 0.11, "windspeed_ms": 2.11},
    {"date": "2026-04-02", "solar_irradiance_kwh_m2": 7.18, "temp_c": 24.9, "precipitation_mm": 0.01, "windspeed_ms": 2.45},
    {"date": "2026-04-03", "solar_irradiance_kwh_m2": 7.13, "temp_c": 25.3, "precipitation_mm": 0.17, "windspeed_ms": 2.23}
  ],
  "source": "NASA POWER",
  "data_retrieved_at": "2026-04-20T00:12:13Z"
}
```

### `get_air_quality` — Open-Meteo Air Quality

Fills the gap that AQICN left when it was removed in v0.1.8. No auth, reliable PH coverage.

```
$ get_air_quality(location="Manila")
```

```json
{
  "location": "Manila",
  "latitude": 14.5995,
  "longitude": 120.9842,
  "measured_at": "2026-04-20T08:00:00Z",
  "pm2_5": 24.8,
  "pm10": 34.3,
  "carbon_monoxide": 521.0,
  "nitrogen_dioxide": 11.6,
  "sulphur_dioxide": 15.5,
  "ozone": 81.0,
  "european_aqi": 65,
  "us_aqi": 91,
  "aqi_category": "Moderate",
  "source": "Open-Meteo Air Quality",
  "data_retrieved_at": "2026-04-20T00:12:13Z"
}
```

### `get_vegetation_index` — NASA MODIS via ORNL DAAC

NDVI + EVI at 250m, 16-day composites. MOD13Q1 product. Useful for monitoring crops, droughts, and deforestation. Here — a rice-bowl pixel in Nueva Ecija going through the growing cycle:

```
$ get_vegetation_index(latitude=15.58, longitude=121.0,
                       start_date="2026-01-01", end_date="2026-04-18")
```

```json
{
  "latitude": 15.58,
  "longitude": 121.0,
  "product": "MOD13Q1",
  "band": "NDVI+EVI (250m, 16-day composite)",
  "samples": [
    {"composite_date": "2026-01-01", "ndvi": 0.708, "evi": 0.343},
    {"composite_date": "2026-01-17", "ndvi": 0.856, "evi": 0.582},
    {"composite_date": "2026-02-02", "ndvi": 0.898, "evi": 0.703}
  ],
  "source": "NASA MODIS via ORNL DAAC",
  "data_retrieved_at": "2026-04-20T00:12:17Z"
}
```

### `get_usgs_earthquakes_ph` — USGS FDSN

Global-network seismic catalogue, filtered to the PH bounding box. Useful for cross-validating PHIVOLCS local magnitudes against USGS Mww/Mwc solutions.

```
$ get_usgs_earthquakes_ph(min_magnitude=5.0, limit=10)
```

```json
[
  {
    "datetime_utc": "2026-04-06T07:22:42Z",
    "magnitude": 5.2,
    "magnitude_type": "mww",
    "depth_km": 10.0,
    "latitude": 10.8435,
    "longitude": 123.8752,
    "place": "0 km S of Tabonok, Philippines",
    "usgs_event_id": "us6000sn00",
    "felt_reports": 37,
    "tsunami": false,
    "url": "https://earthquake.usgs.gov/earthquakes/eventpage/us6000sn00",
    "source": "USGS FDSN"
  },
  {
    "datetime_utc": "2026-04-04T10:34:28Z",
    "magnitude": 6.0,
    "magnitude_type": "mww",
    "depth_km": 67.0,
    "latitude": 4.8733,
    "longitude": 126.1392,
    "place": "95 km SE of Sarangani, Philippines",
    "usgs_event_id": "us6000smj4",
    "tsunami": false,
    "url": "https://earthquake.usgs.gov/earthquakes/eventpage/us6000smj4",
    "source": "USGS FDSN"
  }
]
```

### `get_historical_typhoons_ph` — NOAA IBTrACS

Every typhoon that has passed through the Philippine Area of Responsibility, from the authoritative IBTrACS track archive. Aggregates track points per storm, falls back across agency wind/pressure solutions (WMO → JTWC → JMA) to populate peak intensity.

```
$ get_historical_typhoons_ph(limit=3)
```

```json
[
  {
    "sid": "2025329N10124",
    "name": "KOTO",
    "season": 2025,
    "basin": "WP",
    "max_wind_kt": 80.0,
    "min_pressure_mb": 975.0,
    "start_time_utc": "2025-11-24T18:00:00Z",
    "end_time_utc": "2025-12-03T00:00:00Z",
    "track_points": 67,
    "passed_within_par": true,
    "source": "NOAA IBTrACS"
  },
  {
    "sid": "2025308N10143",
    "name": "FUNG-WONG",
    "season": 2025,
    "basin": "WP",
    "max_wind_kt": 115.0,
    "min_pressure_mb": 943.0,
    "track_points": 75,
    "passed_within_par": true,
    "source": "NOAA IBTrACS"
  },
  {
    "sid": "2025305N10138",
    "name": "KALMAEGI",
    "season": 2025,
    "basin": "WP",
    "max_wind_kt": 115.0,
    "min_pressure_mb": 948.0,
    "track_points": 43,
    "passed_within_par": true,
    "source": "NOAA IBTrACS"
  }
]
```

### `get_world_bank_indicator` — World Bank Open Data

Any World Bank indicator for the Philippines. Accepts the canonical WB code (e.g. `NY.GDP.MKTP.CD`) or a friendly alias from a curated list (`gdp`, `gdp_per_capita`, `poverty_ratio`, `inflation`, `urban_population_pct`, `internet_users_pct`, `gini`, `tax_revenue_pct_gdp`, etc. — 25 aliases in total).

```
$ get_world_bank_indicator(indicator="gdp", per_page=10)
```

```json
{
  "indicator_id": "NY.GDP.MKTP.CD",
  "indicator_name": "GDP (current US$)",
  "country": "Philippines",
  "country_iso3": "PHL",
  "observations": [
    {"year": 2024, "value": 461617509782.36, "unit": ""},
    {"year": 2023, "value": 437055627244.42, "unit": ""},
    {"year": 2022, "value": 404353369604.63, "unit": ""}
  ],
  "source": "World Bank Open Data"
}
```

## Changelog

Full detail in [CHANGELOG.md](CHANGELOG.md) (Keep a Changelog format). Version
summary:

| Version | Date | Highlights |
|---|---|---|
| **0.4.0** | 2026-05-18 | PSA economy expansion (`get_inflation_stats` regional CPI, `get_labor_stats` LFS rates, `get_health_indicators`) + the auto-stitch `get_area_profile` one-call cross-source context layer with per-capita normalization. 25 → 29 tools. Browse-discovery preserved (no hardcoded `.px`); per-table vintage. |
| 0.3.1 | 2026-05-01 | Correctness pass: no fabricated PAGASA advisories, stoplisted hazard tokens, "City of Manila" coordinate bridge, province/agency-alias infra search. |
| 0.3.0 | 2026-04-27 | PH Accountability layer: PSGC resolver, infra spending search, cross-source anomaly indicators. 17 → 25 tools. |
| 0.2.0 | 2026-04-19 | Six no-auth scientific/open-data sources (NASA POWER, Open-Meteo AQ, MODIS, USGS, IBTrACS, World Bank). 11 → 17 tools. |
| 0.1.x | — | Initial release: PHIVOLCS, PAGASA, PhilGEPS, PSA. 11 tools. |

Per-version detail is also inlined above under each **What's new in vX.Y.Z**
section.

## Roadmap

Requires deeper reverse-engineering than this release — shipped separately when ready:

- `get_active_disasters` / `get_situational_report` via NDRRMC monitoring dashboard (intermittent availability)
- `assess_hazard(lat, lng)` via HazardHunterPH — the top-level GeoRisk ArcGIS catalog is public, but the individual PHIVOLCS/MGB hazard layers (flood, landslide, liquefaction) return `"code": 499, "message": "Token Required"`. Needs a different integration strategy
- `get_flood_layers(lat, lng)` via Project NOAH — current site is an Angular SPA whose XHR surface needs browser-level capture. Deferred
- DPWH Transparency portal direct integration — `transparency.dpwh.gov.ph` and `api.transparency.dpwh.gov.ph` currently sit behind a Cloudflare bot challenge that returns 403 to every non-browser client regardless of User-Agent. v0.3.0 sidesteps this by sourcing infra notices from the open PhilGEPS listing instead; if/when DPWH lifts the block, `sources/infra.py` is the single integration point to swap in.

## Prior art

Other Philippine civic-data MCP servers, each single-dataset:

- [GodModeArch/psgc-mcp](https://glama.ai/mcp/servers/@GodModeArch/psgc-mcp) — PSA Philippine Standard Geographic Code (administrative hierarchy)
- [GodModeArch/ph-holidays-mcp](https://glama.ai/mcp/servers/GodModeArch/ph-holidays-mcp) — Philippine national holidays from the Official Gazette
- [GodModeArch/lts-mcp](https://glama.ai/mcp/servers/GodModeArch/lts-mcp) — DHSUD License to Sell registry
- [xiaobenyang-com/Philippine-Geocoding](https://glama.ai/mcp/servers/@xiaobenyang-com/Philippine-Geocoding) — PSGC geocoding
- [darwinphi/ph-schools-mcp-server](https://github.com/darwinphi/ph-schools-mcp-server) — DepEd schools masterlist

Non-MCP libraries that inspired this project:

- [panukatan/lindol](https://github.com/panukatan/lindol) — R package for PHIVOLCS earthquakes
- [pagasa-parser](https://github.com/pagasa-parser) — JS org for PAGASA data parsing

`ph-civic-data-mcp` is the first MCP that unifies multiple Philippine civic-data sources (PHIVOLCS, PAGASA, PhilGEPS, PSA) behind one interface, and the first to expose hazards, weather, procurement, statistical data, solar/climate, air quality, satellite vegetation indices, and macro indicators (v0.2.0) as MCP tools. Credit to all of the above.

## License

MIT. Xavier Puspus. Not affiliated with PHIVOLCS, PAGASA, PhilGEPS, or PSA.

## Contributing

Issues and PRs welcome at [github.com/xmpuspus/ph-civic-data-mcp](https://github.com/xmpuspus/ph-civic-data-mcp).
