AI-ready Fintech API Platform

Integrate Indian fintech APIs faster with AI coding agents

Connect Claude Code, Cursor, Codex, Copilot, or any MCP-compatible agent to Eko Platform Services. Your agent gets the right API context, HMAC signing logic, sandbox flows, and multi-step payment and verification recipes — without digging through docs.

Install for your agent Browse Toolkit

AI Vibe Coding
Claude Code
Opus 4.8 with high effort
❯ Add PAN verification to my onboarding form
● I'll wire this up with the EPS MCP.
● eps · search("verify PAN")
⎿ pan-lite, pan-advanced, bulk-pan-verify
● eps · get_signing_snippet("node")
⎿ HMAC-SHA256 signing ready
● Writing src/kyc/verify-pan.ts …
⎿ Updated 1 file (+34 -0)
✓ PAN verification wired — signed EPS call, name-match on

The problem

API integrations break when agents don’t understand the real workflow

Generic API specs help with endpoints. Real fintech integrations need more: HMAC signing, access-key handling, sandbox headers, multi-step flows, OTP steps, sender/recipient state, and error-specific next actions.

Auth mistakes

HMAC signatures, timestamps, and secret-key handling are easy to get wrong.

Broken flows

Payment and banking APIs often require multiple dependent calls, not one endpoint.

Context overload

Dumping full API docs into an LLM wastes tokens and still misses the right flow.

Why agents get it right

EPS gives your agent integration-grade context

The EPS MCP server exposes only the context your agent needs at each step: endpoint discovery, request schemas, signing snippets, SDK examples, sandbox flows, and recipe-aware error handling.

Generate signed requests correctly

EPS requests are HMAC-signed with your secret-key — the exact step generic OpenAPI tooling gets wrong. Our context teaches your agent the signing scheme, so it works on the first call instead of the fifth.

Smaller context, better answers

A lazy, tiered MCP: your agent lists endpoints, then drills into just the one it needs. It never dumps the whole spec into the context window, so you keep tokens for the actual task.

Works with your coding agent

Use EPS with Claude Code, Cursor, Codex, GitHub Copilot, OpenCode, Aider, Zed, JetBrains AI, and other MCP-compatible tools.

Install

Connect EPS to your coding agent

Install the EPS MCP server and skills into your coding agent. Claude Code and Codex have a two-step native plugin install; every other agent wires the MCP directly. Pick your tool below.

Install for your agent — Claude Code, Codex, Cursor, Copilot, Antigravity, Zed, Kiro, Aider, JetBrains…

Claude Code

MCP

Context pack

Install the plugin · skills + /eps

Step 1 — add the marketplace

/plugin marketplace add ekoindia/eps-platform

Step 2 — install the plugin

/plugin install eps@ekoindia

Run both inside the Claude Code prompt. Installs the skills and the /eps command, and wires the eps MCP automatically. The manual MCP command below is only needed if you skip the plugin.

MCP install command

claude mcp add eps --scope project -- npx -y @ekoindia/eps-context-mcp@latest

`--scope project` writes a shared `.mcp.json` committed with the repo. Use `--scope user` for every project on this machine, or drop the flag for local scope (private to this checkout, not shared).

Context pack

CLAUDE.md./CLAUDE.md

The toolkit

Everything your agent needs

Generated from one API source of truth, so every artifact stays in lock-step with the live platform.

Local EPS MCP server

Your coding agent integrates and tests EPS APIs end-to-end on its own — discovering endpoints, wiring correct HMAC auth, and running signed sandbox calls. Zero hosting, zero secrets; the tiered context server it talks to.

npx -y @ekoindia/eps-context-mcp@latest

HMAC-signed SDKs

Backend-only SDKs with HMAC signing baked in. Keep your access_key server-side.

npm i @ekoindia/eps-sdk

composer require ekoindia/eps-sdk

Offline fintech API mock server

Pairs with the MCP for offline testing — replays golden sample responses with recipe-aware error branching, so your agent develops and tests without ever touching the live API.

npx -y @ekoindia/eps-mock-server

OpenAPI 3.1 spec

Canonical OpenAPI v3.1 specification of all EPS APIs in JSON format — useful for API clients, code generators, validation tools, and other OpenAPI-compatible workflows.

openapi.json

Postman sandbox collection

Every EPS endpoint, ready to import and run against the sandbox.

eps.postman_collection.json

EPS context pack (fallback)

An EPS section to append to your existing AGENTS.md — for agents without MCP or skills support (Aider, JetBrains AI, …).

AGENTS.md CLAUDE.md .cursorrules copilot-instructions.md

Multi-step recipes

Real flows, not just endpoints

EPS encodes the conditional, multi-call runbooks an integration actually needs — including the error branches.

DMT — Send Money

Full domestic money transfer flow: look up the sender, onboard them if new, add the recipient, then send an OTP-verified transfer.

1
dmt-get-sender
Check whether the customer is already a registered DMT sender.
463 → dmt-onboard-sender · Sender not found — onboard them before continuing.
2
dmt-onboard-sender
Register a new sender when Get Sender returned 463.
3
dmt-add-recipient
Add the beneficiary the sender wants to transfer to.
4
dmt-send-otp
Trigger the transaction OTP sent to the sender.
5
dmt-initiate-transfer
Submit the OTP-verified transfer to complete the flow.
0 → complete

AePS — Cash Withdrawal

Aadhaar-enabled cash withdrawal: one-time agent activation, daily 2FA, then the biometric withdrawal.

1
aeps-activate-fingpay
One-time activation of AePS Fingpay for the agent.
2
aeps-daily-auth
Daily two-factor authentication required before transacting.
3
aeps-cash-withdrawal
Perform the biometric Aadhaar-enabled cash withdrawal.
0 → complete

Start building with EPS in your AI coding agent

Wire EPS into your coding agent — ship your first signed EPS call today.

Open API Docs Install by Harness