Eko Platform Services

The AI-ready developer platform for Eko Platform Services — India's fintech APIs (AePS, DMT, BBPS, PPI wallets, KYC & verification, agent banking).

A single source-of-truth data layer describes every EPS API once. From it, the build generates everything a developer — or their AI coding agent — needs to integrate: backend SDKs, an MCP server, drop-in AI context packs, an offline mock server, OpenAPI + Postman, LLM-readable docs, and the marketing + developer-docs website itself. Edit the data, and every artifact regenerates in sync.

Why this exists

• One source of truth, zero drift. Specs, SDKs, MCP, docs, OpenAPI and code samples all derive from the same data layer (src/lib/data/). No hand-maintained copies that fall out of date.
• AI-ready by design. Ships an MCP server plus drop-in context packs (CLAUDE.md, AGENTS.md, .cursorrules, copilot-instructions.md) so agents integrate EPS correctly — including the tricky parts: per-request HMAC-SHA256 signing and multi-step recipes (e.g. the DMT 463 → onboard sender branch).
• Backend-safe SDKs. JS and PHP SDKs are generated from an SDK surface with signing built in and cross-language golden-vector conformance tests. Coverage is a growing subset — see the API coverage roadmap.
• Open & offline. The MCP runs locally with a baked-in bundle (no hosted service, no secrets); an offline mock server replays recorded fixtures for tests.

Use EPS

# AI agent context (any MCP-capable harness)
npx -y @ekoindia/eps-context-mcp

# Backend SDKs (server-side only — signing key must stay secret)
npm install @ekoindia/eps-sdk          # Node.js
composer require ekoindia/eps-sdk      # PHP 8.1+

# Offline mock server for local dev & tests
npx -y @ekoindia/eps-mock-server

Install the EPS eps plugin (MCP + skills + /eps command) into your coding agent — Claude Code and Codex have a native two-step plugin install, every other agent wires the MCP directly. See the per-agent install matrix at the site's /ai hub, and point any agent at /llms.txt for the full machine-readable index.

Project structure

src/
  lib/
    data/        # ← source-of-truth data layer (specs, products, recipes, industries, solutions, pricing)
    agent/       # generators: agent bundle, fixtures, install matrix, context packs
    sdk/         # generators: SDK surface, Postman collection
    openapi/     # generator: OpenAPI 3.1 document
    markdown/    # generators: /llms.txt, /index.md, /products.md, per-page markdown
    config/      # site-wide config constants (site URL, API endpoints)
  components/ hooks/ pages/ assets/   # the React website
  App.tsx / AppServer.tsx / entry-server.tsx / main.tsx   # SPA + SSG entry points
ssg/             # custom Vite plugin for static pre-rendering
public/          # static files served as-is (robots.txt, _redirects)
docs/            # detailed documentation (index below)
vite-plugin-generate-*.ts   # build-time emitters (openapi, agent bundle, markdown, xlsx)

packages/        # published artifacts (each baked from the build output):
  eps-context-mcp/   # @ekoindia/eps-context-mcp — local docs/context MCP server (npm)
  eps-transact-mcp/  # @ekoindia/eps-transact-mcp — transactional MCP server: remote HTTP + local stdio (private)
  sdk-js/            # @ekoindia/eps-sdk — Node.js backend SDK (npm)
  eps-mock-server/   # @ekoindia/eps-mock-server — offline mock server (npm)
  sdk-php/           # ekoindia/eps-sdk — PHP backend SDK (Composer/Packagist)
  claude-plugin-eps/ # agent plugin: dev-time EPS context (manifest + skills + /eps command; not an npm package)

dist/                  # build output (generated, gitignored)
packages/*/data/       # baked artifacts copied here at build time (generated, gitignored — never hand-edit)

That's 3 npm packages, 1 Composer package, and 1 agent plugin bundle (the eps context plugin, installed via each agent's native plugin manager or MCP config — see /ai). The transactional MCP server ships as its own @ekoindia/eps-transact-mcp package (hosted + local stdio); it is not a coding-agent plugin.

Caution

The packages/*/data/ files are generated outputs — they may be present in your working tree but are gitignored and must never be hand-edited or committed.

Data & configuration

Source-of-truth data (src/lib/data/):

File	Role
api-specs.ts	Technical REST specs, one entry per API (deltas only)
api-auth.ts / api-error-codes.ts / api-specs-common.ts	Shared auth headers, error catalog, common params/envelopes + resolvers
api-products.ts	API product registry + metadata
api-product-pages.ts	Product-page marketing content (hero, features, FAQs, SEO) — no technical data
api-recipes.ts	Multi-step integration recipes (e.g. DMT onboard-then-transfer) — drives /recipe + /recipe/<slug> and the agent bundle
recipe-graph.ts	Resolves a recipe into steps + edges; shared by the HTML stepper and the .md twin so they can't drift
industries.ts / solutions.ts	Industry pages + solution/pack definitions
api-pricing.ts / payments-pricing.ts	Pricing data (drives pricing page + xlsx calculator)
docs-registry.ts	Merges MDX guides + API endpoints into the docs nav

Config files (root): package.json (npm workspaces), vite.config.ts, vitest.config.ts, tsconfig*.json, eslint.config.js, components.json, the vite-plugin-generate-*.ts emitters, and per-platform deploy rewrites (vercel.json, netlify.toml, nginx.conf, .htaccess).

Tip

A full index of every configuration file lives in docs/configuration.md.

How it works

src/lib/data/ ──▶ npm run build
                   ├─ vite build   → dist/ (website, openapi.json, /agent/*.json + per-API & per-topic slices,
                   │                  context packs, /llms.txt, /index.md, /products.md, pricing xlsx)
                   └─ bake:all     → copies the relevant artifacts into each packages/*/data/
                                     (then each package builds & publishes from there)

npm run build runs the Vite build and then invokes bake:all automatically — you don't run them separately.

Develop this repo

Setup (Node >=20):

npm install
npm run dev          # local dev server
npm run build        # full build + bake (generates all artifacts)
npm test             # vitest run
npm run lint         # eslint

npm install, npm run build, npm run lint, and npm test all run without any credentials — no API keys are needed to build the site or run the suite. Sandbox credentials are only required to exercise the /docs "Try it" console live: copy .env.example → .env.local and fill in your own Eko UAT/sandbox keys. Those values are read only during vite dev (gated by import.meta.env.DEV), are never bundled into production, and the access key is used purely for in-browser HMAC signing. See CONTRIBUTING.md for the full contributor workflow.

How to add things:

Add a…	Edit	Guide
API product	api-products.ts + api-product-pages.ts, then add it to ProductsSection.tsx tabs	docs/configuration.md
API spec	api-specs.ts (set productId)	docs/api-specs.md
Dev-docs API reference	(auto from api-specs.ts)	docs/developer-docs/api-documentation.md
Dev-docs guide (MDX prose)	add MDX guide	docs/developer-docs/mdx-guides.md
Industry / Solution pack	industries.ts / solutions.ts	docs/industry-and-packs-plan.md
AI artifact / context pack	src/lib/agent/ generators	docs/ai-agent-platform.md

Gotchas:

• packages/*/data/ and dist/ are generated by the build/bake:all — never hand-edit, never commit (gitignored). If a preview shows changes there, that's build output, not source.
• The site builds to static pre-rendered HTML (SSG) for every known route; unknown routes fall back to the SPA shell (__spa-fallback.html) via each platform's catch-all rewrite. See docs/static-page-generation.md and docs/ssg-hydration.md.
• After a redeploy, stale JS chunks auto-reload — see docs/chunk-error-auto-reload.md.
• Deploy: build locally, then deploy dist/ to any of Vercel / Netlify / Apache / Nginx — each reads only its own rewrite file (vercel.json / netlify.toml / .htaccess / nginx.conf), so they don't conflict.

Documentation

Platform & AI

• AI agent platform
• Transactional MCP server
• API coverage roadmap

API & data model

• API technical specifications
• Markdown / LLM content generation
• Developer-docs portal architecture
• Industries & solution packs plan

SDKs & release

• SDK golden-vector conformance
• Releasing the agent packages

Site build & ops

• Configuration — every config file
• Static page generation (SSG) · hydration rules
• Pricing calculator · Command palette (⌘K) · Stale-chunk auto-reload

Tech stack

Vite · React · TypeScript · Tailwind CSS · shadcn-ui · npm workspaces.

Contributing & security

• CONTRIBUTING.md — dev setup, branch/commit conventions, PR workflow.
• SECURITY.md — how to report a vulnerability (please do not open a public issue for security reports).
• CODE_OF_CONDUCT.md — community standards.

License

Source code is released under the MIT License. The MIT grant covers the code only. The "Eko" name, logos, and brand assets are trademarks of their owner and are not licensed for reuse. Marketing copy, page content, and design assets under src/ and public/ may be subject to separate terms — see LICENSE for the exact scope before reusing non-code material.