Developer tools for Nigeria's National Digital Alphanumeric Postcode System (NDAPS), the building-level postcode NIPOST launched in October 2026: libraries for Rust, Python, JavaScript, and Java, MCP servers for AI assistants on PyPI and npm, and a resolver that turns described addresses into postcodes.
A postcode has 11 characters in five segments, written EK-01-A03-FK-01: state, LGA, district, area and building unit.
| Package | What it does | Install | Source |
|---|---|---|---|
ng-postcode (Rust) |
Parse, validate and format codes offline; client for the postcode.gov.ng API | cargo add ng-postcode |
src/, docs |
ng-postcode (Python) |
The same behaviour, with sync and async clients | pip install ng-postcode |
python/ |
ng-postcode-js |
The same behaviour for JavaScript and TypeScript, with a fetch client |
npm install ng-postcode-js |
js/packages/ng-postcode-js |
ng-postcode and ng-postcode-api (Java) |
The same behaviour for Java 17 and later; the core has no dependencies | io.github.adeniyikayodee:ng-postcode-api |
java/ |
ng-postcode-mcp |
MCP server: validate, look up, autocomplete, find by location, resolve addresses | uvx ng-postcode-mcp |
mcp/ |
ng-postcode-mcp (npm) |
The same server for Node, without address resolution | npx ng-postcode-mcp |
js/packages/ng-postcode-mcp |
ng-postcode-polars |
Polars expressions that parse, format and group codes by place, compiled from the Rust core | pip install ng-postcode-polars |
native/polars |
ng_postcode (PostgreSQL) |
The postcode as native 8-byte types with index support, compiled from the Rust core | Package from Releases, then CREATE EXTENSION ng_postcode |
native/pg |
ng_postcode (DuckDB) |
The same functions as a DuckDB extension, compiled from the Rust core | Build from ng-postcode-duckdb | ng-postcode-duckdb |
ng-address-resolver |
Resolve free-text addresses to postcodes, only as precisely as the evidence allows (pre-alpha) | pip install ng-address-resolver |
agent/ |
Each package has its own README with full usage.
For databases, sql/ holds the postcode as a PostgreSQL domain and as DuckDB macros, in plain SQL that runs the same shared cases.
Two names are easy to get wrong. On npm the library is ng-postcode-js, because ng-postcode there is another project. ng-address-resolver is imported as ng_address.
For online stores, ng-postcode-woocommerce is a WooCommerce plugin built on the same rules and the same shared test cases: it checks and tidies the postcode at checkout, finds it from the customer's location, and makes shipping zones match on postcode prefixes.
Parsing and validation are offline and need no key.
Python
from ng_postcode import Postcode, parse
match parse("ek 01 a03 fk 01"):
case Postcode() as code:
print(code, code.compact) # EK-01-A03-FK-01 EK01A03FK01
case error:
print(error) # e.g. "invalid lga segment"JavaScript and TypeScript
import { Postcode, parse } from "ng-postcode-js";
const code = parse("ek 01 a03 fk 01");
if (code instanceof Postcode) {
console.log(String(code), code.compact); // EK-01-A03-FK-01 EK01A03FK01
} else {
console.log(String(code)); // e.g. "invalid lga segment"
}Java
import io.github.adeniyikayodee.ngpostcode.Postcode;
if (Postcode.parse("ek 01 a03 fk 01") instanceof Postcode code) {
System.out.println(code + " " + code.compact()); // EK-01-A03-FK-01 EK01A03FK01
}Rust
use ng_postcode::{Postcode, Segment};
let code: Postcode = "ek 01 a03 fk 01".parse()?;
assert_eq!(code.to_string(), "EK-01-A03-FK-01");
assert_eq!(code.prefix(Segment::Area), "EK-01-A03-FK");AI assistants
The MCP server works with any MCP client. It runs over stdio as uvx ng-postcode-mcp, with the API key in the environment. Most clients take this entry in their MCP settings:
{
"mcpServers": {
"ng-postcode": {
"command": "uvx",
"args": ["ng-postcode-mcp"],
"env": { "NG_POSTCODE_API_KEY": "nipost_live_..." }
}
}
}To run it with Node, use "command": "npx" and "args": ["-y", "ng-postcode-mcp"]; that edition has every tool except address resolution. Per-client steps for Cursor, VS Code, Codex, Claude and others are in the MCP server README. Validation works without a key; the other tools need one from the NIPOST developer dashboard. The server is listed in the MCP Registry as io.github.Adeniyikayodee/ng-postcode.
PostgreSQL
Load the plain SQL file with psql -f sql/postgres.sql. It needs no extension, so it runs on managed services.
CREATE TABLE customers (
id bigint PRIMARY KEY,
postcode ng_postcode -- refuses anything that is not a compact code
);
INSERT INTO customers VALUES (1, ng_postcode_parse('ek 01 a03 fk 01'));
SELECT ng_postcode_format(ng_postcode_truncate(postcode, 'district')) AS district, count(*)
FROM customers
GROUP BY 1;
-- EK-01-A03 | 1On a server where you can install an extension, the native extension has the same functions and stores a code in 8 bytes. Unpack the package for your PostgreSQL version from Releases, then run CREATE EXTENSION ng_postcode.
DuckDB
Load the same functions with duckdb my.db < sql/duckdb.sql.
SELECT ng_postcode_parse('ek 01 a03 fk 01');
-- EK01A03FK01ng-postcode-duckdb provides them as a compiled extension.
Polars
import polars as pl
import ng_postcode_polars # pip install ng-postcode-polars
frame = pl.DataFrame({"postcode": ["ek 01 a03 fk 01", "EK-01-A03-ZZ-14", "100001"]})
print(frame.with_columns(code=pl.col("postcode").ng.parse(), ok=pl.col("postcode").ng.is_valid()))
# ek 01 a03 fk 01 EK01A03FK01 true
# EK-01-A03-ZZ-14 EK01A03ZZ14 true
# 100001 null falseA CSV file
pip install ng-postcode
ng-postcode --column postcode customers.csv --out checked.csvThe output keeps every row and adds the compact code, a status of valid, suggested, invalid, or empty, a suggested correction, and the reason for each invalid row.
Language models trained before October 2026 expect a six-digit Nigerian postcode. To point a coding agent at the current format, paste this into your project's AGENTS.md, CLAUDE.md, or editor rules:
Nigerian postcodes: since 1 October 2026 a building's postcode is NIPOST's 11-character
digital postcode (NDAPS), such as EK-01-A03-FK-01, and a six-digit pattern does not match it.
Validate and format it with ng-postcode (pip, cargo), ng-postcode-js (npm), or
io.github.adeniyikayodee:ng-postcode (Maven). On npm, ng-postcode is a different project.
Reference: https://adeniyikayodee.github.io/ng-postcode/llms.txt
| Segment | Example | Shape |
|---|---|---|
| State | EK |
2 letters |
| LGA | 01 |
2 digits, 01 to 99 |
| District | A03 |
3 letters or digits |
| Area | FK |
2 letters |
| Building unit | 01 |
2 digits, 01 to 99 |
| Form | Example | Use it to |
|---|---|---|
| Canonical | EK-01-A03-FK-01 |
Write a code and pass it between systems |
| Spaced | EK 01 A03 FK 01 |
Show a code to people |
| Compact | EK01A03FK01 |
Store and compare codes |
Input may be hyphenated, spaced or compact, in either case. The compact form matches ^[A-Z]{2}(0[1-9]|[1-9][0-9])[A-Z0-9]{3}[A-Z]{2}(0[1-9]|[1-9][0-9])$. A well-formed code is not necessarily assigned to a building; only the NIPOST API can confirm that.
docs/schemas/postcode-reference.schema.json defines a small JSON object for handing a location from one system or AI agent to another: the code, its level (state to building), and optionally a confidence and whether NIPOST confirmed it. A resolved or partial resolve_address answer already fits it.
For a whole address, docs/address-record.md wraps the reference in an address record and maps it onto ISO 20022, FHIR, schema.org, vCard, and UPU's address layout for Nigeria.
- One behaviour, four languages: Rust, Python, JavaScript and Java all run the cases in
spec/, so they cannot drift apart. The Node MCP server registers its tools fromspec/mcp.json, which the Python server generates. - Offline first: parsing and validation never touch the network. API access is a separate, optional layer.
- Errors are values: expected failures, such as a malformed code or a rejected API key, come back as return values.
- Careful with money and guesses: the MCP server caps lookups at the free level unless told otherwise, and never corrects a mistyped code into a paid call. The resolver gives an area or district code when that is all the evidence supports.
- The NIPOST API needs a key for every endpoint, from the developer dashboard. Offline validation needs nothing.
- The API layer is tested against responses captured from the live API with a level 1 key, kept in
spec/responses.json. Runscripts/live_check.pywith your own key to repeat the comparison. Lookup levels 2 and up need a higher-access key and are tested only against NIPOST's documented examples. - The resolver and the
resolve_addresstool are pre-release. They work against the live API, but their accuracy on real addresses is unmeasured. Described addresses need a geocoder you run or pay for; text alone rarely identifies a building, so ask users for a location pin when the exact building matters. - Notes for integrators on how the live API behaves, each with the date it was observed, are in
docs/api-differences.md.
See CONTRIBUTING.md for the shared spec, how to add a language, and the commit standards.
cargo test --all-features # Rust
cd python && uv run --group dev pytest # Python library
cd mcp && uv run --group dev pytest # MCP server
cd agent && uv run --group dev pytest # resolver
cd js && npm ci && npm test # JavaScript library and Node MCP server
cd java && ./mvnw verify # Java librariesCI runs formatting, linting, type checks and tests for every package. Releases publish from tags (v* to crates.io; py-v*, mcp-v* and agent-v* to PyPI and the MCP Registry; js-v* and js-mcp-v* to npm) through trusted publishing, so no tokens are stored. Maven Central has no trusted publishing, so java-v* uses a token and a signing key held as environment secrets, and the upload is published by hand. An mcp-v* release also creates a GitHub release, which Glama rebuilds its listing from, and republishes the Smithery bundle with a key held the same way.
MIT