MCP · REST API · Multi-Tenant

Build with
Abhi's rewards engine

Embed AI-powered credit card recommendations into your fintech, bank app, or AI agent. MCP-native with a simple REST fallback — deployed and running today.

3,500+ cards catalogued
JSON-RPC 2.0 MCP server
Raw rewards data never exposed
Per-tenant persona & tone

Four tools.
One API key.

Everything your app needs to give users smart, real-time credit card advice — without touching raw rewards data.

🔍

Card Search

Search 3,500+ credit cards by name or issuer. Returns card names and issuers — safe for any UI, no licensing restrictions.

search_cards
🎯

Smart Recommendations

Tell Abhi where your user is shopping. Get back a plain-language recommendation with the best card from their wallet and why.

recommend_card
💳

Wallet Management

Read and manage a user's registered card wallet. Add/remove cards with built-in cross-channel WhatsApp confirmation for safety.

get_wallet · add_card
🤖

Conversational Chat

Full multi-turn conversation with Abhi's rewards agent. Ask anything — benefits, travel perks, annual fees — in plain language.

chat

Up and running
in minutes

One API key, one endpoint. Works as a standard REST API or as a native MCP server for AI agents.

01

Get your API key

Contact us to provision a tenant. You get a unique API key shown once — store it securely.

02

Add the header

Include x-api-key: your_key on every request. Phone number is the universal user identity.

03

Call the tools

POST to /mcp as JSON-RPC 2.0, or use the REST shortcuts at /v1/*.

04

Configure your persona

Set tone (casual/formal), response format (text/JSON), emoji use, and optionally filter to your card issuers.

Try the API now

No API key needed. Each browser tab gets its own isolated demo session — add cards, ask questions, see real responses. Session resets after 30 min of inactivity.

POST /mcp · JSON-RPC 2.0
Live
Demo session: #----
Tool
Search Query
Response
// No API key needed — this is a live demo. // Your session is isolated from other users. // Pick a tool and hit Run Demo.

Python sample code

Connect to the MCP server in under 20 lines. Replace your_api_key_here with your key once you have access.

Python · requests
import requests

API_KEY  = "your_api_key_here"
MCP_URL  = "https://abhi-ai.com/mcp"
HEADERS  = {"Content-Type": "application/json", "x-api-key": API_KEY}

def call_tool(name: str, **kwargs) -> str:
    """Call any Abhi MCP tool and return the text response."""
    resp = requests.post(MCP_URL, headers=HEADERS, json={
        "jsonrpc": "2.0",
        "id":      1,
        "method":  "tools/call",
        "params":  {"name": name, "arguments": kwargs},
    })
    resp.raise_for_status()
    return resp.json()["result"]["content"][0]["text"]


# 1. Search the card catalog (no phone needed)
print(call_tool("search_cards", query="Chase Sapphire"))

# 2. Add a card to a user's wallet
print(call_tool("add_card",
    phone="+14155550100",
    card_id="chase-sapphire-preferred",
    card_name="Chase Sapphire Preferred",
))

# 3. View the wallet
print(call_tool("get_wallet", phone="+14155550100"))

# 4. Get a merchant recommendation (firewall: no raw rewards returned)
print(call_tool("recommend_card",
    phone="+14155550100",
    merchant="Whole Foods Market",
))

# 5. Conversational chat
print(call_tool("chat",
    phone="+14155550100",
    message="Which card for a hotel stay in London?",
))

# 6. Remove a card
print(call_tool("remove_card",
    phone="+14155550100",
    card_id="chase-sapphire-preferred",
))

Endpoint reference

All endpoints require x-api-key header. Phone number is mandatory in every request body.

Method Path Description
POST /mcp MCP server — JSON-RPC 2.0. All tools available here. Preferred for AI agent integration.
POST /v1/chat Conversational chat with the rewards agent. Multi-turn, maintains session per phone.
POST /v1/recommend Card recommendation for a merchant or spending category.
🔒 Recommendation firewall — no raw rewards data
POST /v1/cards/search Search the card catalog. Returns card names and issuers only — safe for any use.
GET /v1/wallet Get a user's registered card wallet. Returns card names.
POST /v1/cards/add Add a card to a user's wallet. Queues a WhatsApp confirmation — the card is added only after the user replies YES.
POST /v1/cards/remove Remove a card from a user's wallet.

Native AI agent integration

Point your AI agent at POST /mcp with your API key. All tools speak JSON-RPC 2.0.

search_cards catalog

Search 3,500+ credit cards by name or issuer. Use to help users find a card before adding it to their wallet.

Args: query
Returns: list of {name, issuer, card_id}
get_wallet wallet

Retrieve the list of credit cards a user has registered with Abhi, identified by phone number.

Args: phone
Returns: list of card names
add_card wallet

Queue a card addition for a user. They receive a WhatsApp prompt to confirm — the card is only added on YES.

Args: phone, card_id, card_name
Returns: confirmation queued status
remove_card wallet

Remove a card from a user's wallet immediately. Logged for analytics.

Args: phone, card_id
Returns: success / error
recommend_card 🔒 firewall

Given a merchant or spending category, returns which card in the user's wallet to use and why — in plain language. Raw rewards data never leaves this layer.

Args: phone, merchant
Returns: recommendation text only
chat agent

Full conversational access to Abhi's rewards agent. Supports multi-turn dialogue, card queries, benefit lookups, and wallet management in one call.

Args: phone, message
Returns: agent response text

Request your
API key

Tell us your use case and we'll provision a tenant with your persona, tone, and issuer config — usually same day.

  • Same-day provisioning
  • Custom persona & tone per tenant
  • Issuer filtering (e.g. Chase-only)
  • Full MCP + REST API access
  • 3,500+ cards in the catalog