Nigeria Admin Divisions MCP

Nigeria's 36 states + the FCT and all 774 Local Government Areas, exposed to AI agents as MCP tools — fully offline.

Status: 📦 Offline — bundled dataset, no API key, no network Package: @braynexservices/nigeria-mcp-admin-divisions v0.1.0 · depends on @braynexservices/nigeria-mcp-core Author: Samuel Orie (Founder) · Owner: Braynex Services Ltd · License: MIT © 2026 Braynex Services Ltd — MIT licensed (copyright retained; see NOTICE).

What it does

Returns Nigeria's administrative divisions — the 36 states + the FCT (37 in total) with a per-state Local Government Area count, and the full LGA list for any state. The whole dataset (37 divisions, 774 LGAs) is compiled into the package: every lookup is served locally. No network call, no API key, ever. Public reference data — no PII. Each response carries source: "static" and a fresh retrievedAt.

Tools

list_states

Lists every state (36 + FCT) with its LGA count. Takes no arguments.

| Param | Required | Notes | |---|---|---| | — | — | no arguments |

Example:

// → list_states({})
"Abia (17 LGAs)\nAdamawa (21 LGAs)\nAkwa Ibom (31 LGAs)\n…\nFCT (6 LGAs)"
// structuredContent: {
//   count: 37,                                    // 36 states + FCT
//   states: [{ state:"Abia", lgaCount:17 }, …],   // 37 rows
//   source: "static", retrievedAt
// }

get_lgas

Lists the Local Government Areas of one state. Matching is case-insensitive; use FCT (or Abuja) for the Federal Capital Territory.

| Param | Required | Notes | |---|---|---| | state | ✅ | state name, e.g. Lagos or FCT (case-insensitive; min 2 chars) |

Example:

// → get_lgas({ "state": "Lagos" })
"Lagos — 20 LGAs: Agege, Ajeromi/Ifelodun, Alimosho, …, Eti-Osa, …, Surulere"
// structuredContent: {
//   state: "Lagos",
//   lgas: ["Agege", "Ajeromi/Ifelodun", …, "Eti-Osa", …, "Surulere"],  // 20 items
//   source: "static", retrievedAt
// }

// → get_lgas({ "state": "Atlantis" })
// isError: true — "State \"Atlantis\" not found. Use list_states for the 37 valid names (36 states + FCT)."

An unknown state returns an actionable error — it suggests close spelling matches where it can, and always points back to list_states for the 37 valid names.

Configuration — ADMIN_DIVISIONS_PROVIDER

| Value | Behaviour | |---|---| | static (default — and the only provider) | Offline. Serves the states/LGAs dataset bundled into the package. No network, no key. |

There is nothing to configure. No API key is required — ever — and no network is used. The env var exists only so a future API-backed source could be slotted in without changing callers.

Run it

npm install && npm run build      # from repo root

npm run smoke -w @braynexservices/nigeria-mcp-admin-divisions        # end-to-end (in-memory MCP client↔server)

npm run start:stdio -w @braynexservices/nigeria-mcp-admin-divisions  # serve over stdio

There is no separate "live" mode: the bundled dataset is the source. npm run smoke:live exists for parity across lanes but simply re-runs the same offline provider and prints its counts — no network is ever hit.

MCP Inspector (interactive GUI):

npx @modelcontextprotocol/inspector node lanes/admin-divisions/dist/stdio.js

Claude Desktop / Code — add to your MCP config (needs Node.js ≥ 20):

{
  "mcpServers": {
    "nigeria-admin-divisions": {
      "command": "npx",
      "args": ["-y", "@braynexservices/nigeria-mcp-admin-divisions"]
    }
  }
}

No env block is needed — the lane is offline. Running from a clone instead: "command": "node", "args": ["/path/to/nigeria-mcp/lanes/admin-divisions/dist/stdio.js"]. Then ask, e.g., "How many LGAs are in Kano?" or "List the local governments in Lagos."

Limits & notes

• Fully offline. The dataset (37 divisions, 774 LGAs) is compiled into the package — no network call and no API key at any point. Every row is stamped source: "static" with a fresh retrievedAt.
• Reference data, not a live registry. Counts reflect a compiled dataset snapshot (extracted 2026-07-15), not a real-time government feed.
• State → LGA-name mapping only. Wards and coordinates from the upstream GeoJSON are intentionally dropped; this lane carries state names, their LGA names, and counts.
• Matching: case-insensitive; FCT and Abuja both resolve to the Federal Capital Territory.
• Data: compiled from temikeezy/nigeria-geojson-data (MIT, © 2025 Temi). The full upstream MIT permission notice ships with the package in NOTICE. Three spellings were normalized on extraction: "Federal Capital Territory" → "FCT", "Nassarawa" → the official "Nasarawa", and "Badagary" → the official "Badagry".

Verification

npm run verify (repo root) → build + all lane smokes + eval. This lane: an in-memory MCP client↔server smoke exercises both tools plus the not-found path and asserts the dataset invariants on every run — 37 divisions (36 + FCT), 774 LGAs total, and Lagos includes "Eti-Osa".