Razorpay Agent Studio: India's First AI Payment Platform, Explained
A single person can operate like a team of 100 agents.
That is the exact sentence Harshil Mathur, CEO and co-founder of Razorpay, said at FTX 2026 in Bengaluru on March 12, 2026. And I think it is one of the most important statements made by any fintech founder this year, because Razorpay is not just talking about it. They shipped it.
On that day, Razorpay launched Agent Studio, described as the world's first AI-native agent platform built directly on top of payment infrastructure. Not beside it. Not connected to it via an API. On top of it, inside it, as the layer through which financial operations now happen autonomously.
India has over 10 million businesses on Razorpay, processing more than 1 billion transactions per quarter. The challenge has never been moving money - UPI, cards, and netbanking solved that years ago. The challenge has always been everything that happens around the transaction: recovering the abandoned checkout, fighting the chargeback, retrying the failed subscription, forecasting whether there is enough cash to run payroll on Friday. That work used to need human teams. Now it needs agents.
This post covers exactly what Razorpay Agent Studio is, how each agent works, what Google's search data tells us about who is asking about this and why, and what it means for anyone building on or competing with Razorpay.

What Is Razorpay Agent Studio? The One-Paragraph Answer
Razorpay Agent Studio is a B2B AI agent marketplace and builder platform, launched March 12, 2026, that allows businesses to deploy autonomous AI agents for payment and post-payment operations. It is built on Anthropic's Claude Agent SDK and runs natively inside Razorpay's payment infrastructure, giving each agent direct access to transaction data, settlement records, customer activity signals, and third-party business tools like Shopify, Tally, QuickBooks, WhatsApp, Slack, and Shiprocket.
Unlike traditional payment automation that triggers fixed rules, Razorpay agents observe financial signals continuously, reason over context, and take action on their own. They do not wait to be told. They detect the abandoned cart and initiate outreach. They receive the chargeback and file the response. They see the failed subscription and apply a smarter retry. They spot the cash shortfall coming in 5 days and send an alert before it becomes a crisis.
Why this matters:
Payments are only the execution layer of commerce. Everything around the transaction - disputes, reconciliation, recovery, forecasting - still requires manual teams. Agent Studio is Razorpay's attempt to eliminate that manual overhead entirely.
All 8 Launch Agents Explained: Features, Use Cases, Partners
Razorpay debuted eight production-ready agents at FTX 2026. Each is designed to address one specific, high-friction operation that previously required manual effort. Here is the complete breakdown:

Dispute Responder Agent
Chargebacks are a real problem for Indian merchants. When a customer disputes a transaction, the merchant has a tight window to respond with evidence, and most small businesses miss it simply because no one is monitoring. The Dispute Responder agent automatically collects transaction evidence (payment logs, delivery confirmation, customer activity), compiles an optimised response, and submits it before the deadline. Higher win rates. Zero manual effort.
Subscription Recovery Agent
Failed subscription payments are the silent killer of SaaS and D2C subscription revenue. Card expiry, insufficient balance, bank declines - these happen constantly. The Subscription Recovery Agent analyses the reason for failure, applies intelligent retry logic (not just 'try again in 24 hours'), and when retries are insufficient, triggers a voice call to the customer using ElevenLabs voice synthesis. That is a real person-quality voice call, not a robocall. I think this is the most technically impressive agent in the launch set.
Abandoned Cart Conversion Agent (SuperU + Nugget by Zomato)
Razorpay is launching two variants of this agent, both targeting the same problem: someone starts checkout, then leaves. The SuperU-powered variant re-engages via WhatsApp or email with personalised offers. The Nugget by Zomato variant does the same for Zomato-integrated merchants. Both send a payment link to complete the purchase. The key innovation here is that this is not a generic reminder email - it is contextual outreach based on the specific transaction, the customer's loyalty status, and available discounts.
Cashflow Forecaster Agent
This one is specifically for the 10 million small and medium businesses on Razorpay that do not have a CFO. The Cashflow Forecaster analyses transaction patterns and predicts the merchant's cash position 3-7 days ahead, with specific alerts for payroll risk, payout failures, and balance shortfalls. Getting a WhatsApp message on Monday that says 'you will be below minimum balance for payroll on Thursday - here are three options' is genuinely transformative for a small business owner.
RTO Shield and RTO Insights
Return-to-origin is one of the biggest margin killers for Indian D2C brands. Cash on delivery orders that get returned from bad pincodes or invalid addresses eat into shipping costs with no revenue. RTO Shield uses LLM-based address validation and historical bad-pincode intelligence to block high-risk COD orders before they ship. RTO Insights provides analytics across pincodes, products, and customer segments to identify what is driving returns systematically.
Settlement Insights
The simplest agent in the set, and maybe the one that will have the highest daily active usage. Settlement Insights sends a WhatsApp message every morning with a summary of yesterday's settlements. No dashboard login. No manual checking. Just the number that matters, delivered where you already are.

The Agentic Experience Platform: Onboarding, Dashboard, and Integration
Alongside Agent Studio, Razorpay launched the Agentic Experience Platform - a complete reimagining of how merchants interact with Razorpay itself. It has three distinct capabilities, all powered by the same Claude Agent SDK foundation.
Agentic Onboarding
Merchant onboarding used to take 30-45 minutes. Provide PAN and a website URL. The platform auto-validates identity against CKYC and government infrastructure in real time, auto-detects business category from the website, and eliminates manual form-filling. Result: onboarding in approximately 5 minutes. That is a 6-9x improvement in activation speed, which matters enormously for merchant conversion rates.
Agentic Dashboard
The Agentic Dashboard replaces static data tables with a natural language interface. You can upload a screenshot of your bank statement and ask 'reconcile this with my Razorpay settlements' - the agent extracts UTR numbers, cross-references them against Razorpay records, and flags discrepancies. You can say 'why did this customer's payment fail?' and get an actual answer. This is genuinely new. Most payment dashboards show you the data. This one reasons over it.
Agentic Integration
Harshil Mathur demonstrated live at FTX 2026 that a developer can complete full payment integration in under two minutes using the Agentic Integration layer. The system auto-detects the tech stack (Claude Code, Replit, Emergent, custom frameworks), generates ready-to-paste code, and handles the setup. The old gold standard was 5-hour integration. Two minutes is a category shift.
The Technical Foundation: Anthropic Claude Agent SDK
The choice to build on Anthropic's Claude Agent SDK is not a minor detail. It is the foundational technical decision that explains why Razorpay Agent Studio works the way it does.
Claude was renamed from the Claude Code SDK to the Claude Agent SDK in late 2025, reflecting that it had evolved far beyond coding assistance into a general-purpose agentic runtime. Razorpay's CPO Khilan Haria stated the company evaluated multiple AI providers and chose Claude for its advanced reasoning capabilities, safety-first design philosophy, and suitability for high-stakes financial workflows where errors are costly.
Each Razorpay agent runs inside a Claude reasoning loop: it observes signals from Razorpay's payment data, reasons over what action to take, executes that action via Razorpay's 400+ APIs and integrated third-party tools, and reports results. Agents operate within consent guardrails - Harshil Mathur was specific about this: the agent never sees raw financial data beyond what is necessary for the task, and all actions occur within the merchant's defined permissions.
Irina Ghose, Managing Director of Anthropic India:
"Razorpay's work with Claude shows how AI agents can address real commerce challenges - recovering revenue, resolving disputes, and predicting cash flow. It's a great example of what AI can do when it's embedded into the operating fabric of business."
The no-code agent builder, currently in beta, uses Claude's natural language understanding to let non-technical users define agent behaviour in plain English. Describe the task, select the systems the agent can access, set the guardrails, and deploy. No engineering dependency.
Agentic Commerce: Chatting with Zomato to Order Samosa
Beyond Agent Studio, Razorpay is building what they call agentic commerce - the ability to complete a purchase entirely through a conversational interface inside an existing app, without navigating menus or checkout pages.
Harshil Mathur described it simply at FTX 2026: 'You can open the Zomato app, chat with it and say, hey, I want samosa from this, I want chai from this, and the Zomato app can buy it for you. Completely autonomous just by chatting on it.'
Razorpay is piloting this with Zomato, Swiggy, PVR Inox, Vodafone Idea, Bluestone, and Honasa (The Derma Co). Each pilot embeds Razorpay's payment infrastructure into the conversational layer of these apps, so when the AI assistant recommends a product, the payment happens in the same flow. Discovery, decision, and payment - all inside one conversation.
This is not a feature. This is a redefinition of what checkout means. The implications for conversion rates, average order values, and the entire checkout abandonment problem are significant. I think agentic commerce will become the dominant commerce model for mobile-first markets like India within 3 years.
Razorpay also announced that developers can sell through ChatGPT with zero code using a Razorpay integration - enabling merchants to monetise their products inside the ChatGPT interface from day one.
Who Should Use Razorpay Agent Studio and How to Get Access
Agent Studio is relevant to any business that processes payments on Razorpay and has operational overhead around managing those payments. Specifically:
• E-commerce and D2C brands - deploy the Abandoned Cart Conversion Agent, RTO Shield, RTO Insights, and Settlement Insights immediately
• SaaS and subscription businesses - the Subscription Recovery Agent is the highest-impact agent for reducing involuntary churn
• Any business receiving international or premium card payments - the Dispute Responder Agent pays for itself after the first successful chargeback win
• SMEs managing cash flow manually - the Cashflow Forecaster Agent is genuinely life-changing for founders who are managing payroll from a single checking account
• Developers and technical teams - the MCP server, llms.txt in the developer docs, and Agentic Integration make Razorpay the most AI-developer-friendly payment gateway in India
Access is via early access signup at razorpay.typeform.com/to/hGkU4Jpb. The platform is currently free during the beta phase. Razorpay has indicated a future token-based pricing model based on agent task volume and complexity.
One honest caveat: Agent Studio is days old as of this writing. The agents are production-ready but the overall ecosystem is early. The no-code builder is in beta. The open third-party marketplace for developers has not launched yet. If you are evaluating whether to rely on specific agents for critical operations, request a demo before committing.
The Bigger Picture: What This Means for AI in Indian Fintech
I want to be direct about what I think is actually happening here, because the media coverage has been mostly descriptive and not analytical enough.
Razorpay is not just adding AI features to a payment gateway. They are attempting a category redefinition. A payment gateway moves money. A financial operating system manages the entire commercial operation around money. Agent Studio is the first product layer of that operating system. The agents are the staff.
The global agentic payments movement is accelerating simultaneously. PayPal launched its Agent Toolkit for developers. Visa announced Intelligent Commerce. AWS published its x402 agentic payment protocol. Coinbase's x402 enables AI agents to make micro-payments between machines. Within fintech, a consensus is forming that the next era of commerce will be AI-mediated, not human-navigated.
India is uniquely positioned for this shift. UPI processes 17+ billion transactions per month. Digital payment adoption is mainstream. And 90%+ of businesses are SMEs without dedicated finance teams - exactly the segment that benefits most from agents doing the work of a team.
The risk I see: the quality of agents at launch is promising but unproven at scale. Agent reliability in financial contexts is not the same as agent reliability in writing tasks. A hallucinated chargeback response that loses a dispute costs real money. Razorpay has addressed this with consent guardrails and human-in-the-loop options, but merchant trust will need to be earned through demonstrated performance, not announced through launch events.
That said - the direction is right. The foundation (Claude Agent SDK) is the strongest available. The distribution (10 million merchants on existing Razorpay) is exceptional. If execution matches ambition, Agent Studio will reshape how Indian businesses run their financial operations within 24 months.
Frequently Asked Questions
What is Razorpay Agent Studio?
Razorpay Agent Studio is the world's first AI agent platform built natively on payment infrastructure, launched March 12, 2026 at FTX 2026 in Bengaluru. Built on Anthropic's Claude Agent SDK, it is a B2B marketplace where merchants can deploy pre-built AI agents for dispute management, cart recovery, subscription retry, cashflow forecasting, and RTO reduction - or create custom agents using a no-code builder. Access is via early signup at Razorpay's website; it is currently free in beta.
What is Razorpay AI?
Razorpay AI is the company's umbrella strategy for embedding artificial intelligence across its entire product suite. It includes Agent Studio (autonomous payment agents), the Agentic Experience Platform (AI-native merchant interactions), biometric card authentication with Mastercard, voice-first payments with Gnani.ai and SuperU, and agentic commerce integrations with Zomato, Swiggy, PVR Inox, and Vodafone Idea. All launched at FTX 2026 on March 12, 2026.
What AI model powers Razorpay Agent Studio?
Razorpay Agent Studio is built on Anthropic's Claude Agent SDK (formerly the Claude Code SDK, renamed in late 2025). Claude powers the reasoning, context understanding, and action execution capabilities of every agent on the platform. Razorpay chose Claude over other options specifically for its advanced reasoning and suitability for high-stakes financial workflows where decisions have real monetary consequences.
How many agents does Razorpay Agent Studio have at launch?
Razorpay launched eight agents at FTX 2026: Dispute Responder, Subscription Recovery (with ElevenLabs voice), two variants of Abandoned Cart Conversion (SuperU and Nugget by Zomato), Cashflow Forecaster, RTO Shield, RTO Insights, and Settlement Insights. A No-Code Agent Builder is also available in beta for creating custom agents. An open third-party developer marketplace is planned for a future release.
Is Razorpay Agent Studio available for small businesses?
Yes - small businesses are the primary target. Harshil Mathur's core pitch at FTX 2026 was that small businesses lack the teams to manage post-payment operations that large companies handle automatically. Agent Studio is designed to give a 5-person team the operational capacity of a 100-person finance department. The no-code builder means no developer is needed to create or deploy agents.
Is Razorpay Agent Studio free?
Yes, during the current beta phase. Razorpay has not announced a paid pricing date, but has indicated that a token-based model will be introduced where merchants pay based on the volume and complexity of tasks agents perform. Get on the early access list now to lock in beta access before pricing is introduced.
What is the difference between Razorpay Agent Studio and the Agentic Experience Platform?
Agent Studio is the marketplace and builder for deploying autonomous agents that handle payment operations (disputes, cart recovery, subscriptions, cashflow). The Agentic Experience Platform is a redesign of the merchant experience itself - covering how merchants onboard (5-minute KYC), integrate payments (under 10 minutes), and interact with their dashboard (natural language). They are complementary products launched together at FTX 2026.
What is agentic commerce and how is Razorpay enabling it?
Agentic commerce is a shopping experience where a user can express intent in natural language inside an existing app - no menus, no search, no checkout form - and an AI agent completes the purchase. Razorpay is enabling this with Zomato, Swiggy, PVR Inox, Vodafone Idea, Bluestone, and Honasa by embedding its payment stack into conversational AI interfaces, so payment is a natural part of the conversation rather than a separate step.
Does Razorpay Agent Studio work with Shopify?
Yes. Razorpay Agent Studio integrates with Shopify, WhatsApp, Shiprocket, Slack, Tally, and QuickBooks, giving agents access to cross-platform business data. For e-commerce merchants on Shopify, the Abandoned Cart Conversion Agent, RTO Shield, and Settlement Insights are immediately deployable. Developer integrations also support Claude Code, Replit, and Emergent.
How does Razorpay use AI to prevent payment fraud?
Every transaction on Razorpay is now monitored by an AI security agent that analyses whether a transaction's pattern matches known fraud signatures and blocks suspicious activity before it completes. Harshil Mathur confirmed at FTX 2026 that this agent never sees raw financial data beyond what is required, and all actions occur within the guardrails of the merchant's consent settings.
How does Razorpay for developers work with AI?
Razorpay provides 400+ documented API endpoints, a Model Context Protocol (MCP) server for LLM-agent integration, and an llms.txt in the developer docs for AI coding tools. The Agentic Integration layer auto-detects a developer's tech stack and completes payment setup in under 10 minutes across Claude Code, Replit, Emergent, and standard frameworks.
Who is the CEO of Razorpay?
Harshil Mathur is the co-founder and CEO of Razorpay. He is an IIT Roorkee alumnus who co-founded Razorpay in 2014 with Shashank Kumar. As of 2025, his net worth is approximately $1.04 billion (Hurun India Rich List 2025). He personally announced Agent Studio at FTX 2026.
What is the commission of Razorpay?
Razorpay charges 2% per successful domestic transaction plus 18% GST, making the effective rate approximately 2.36%. International cards, AMEX, Diners, EMI, and corporate cards are charged 3% plus GST. There is no setup fee, no annual maintenance charge, and no minimum transaction commitment. Agent Studio is currently priced separately (free in beta).
Is Razorpay good for freelancers?
Yes. Razorpay supports freelancers and unregistered businesses with payment links, QR codes, no-code payment pages, UPI acceptance, and PAN-only registration with no setup fees. The new Agentic Onboarding reduces account activation to under 5 minutes. Freelancers do not need Agent Studio's operational agents but benefit significantly from the payment link infrastructure and fast onboarding.
Recommended Reads
If you found this useful, these posts from Build Fast with AI go deeper on related topics:
• Top 11 AI-Powered Developer Tools Transforming Workflows in 2025
• AI in 2026: Your Survival Guide to the Fourth Year of Generative AI
• 7 AI Tools That Changed Development — December 2025 Guide
Want to build AI payment agents and automation systems like Razorpay?
Join Build Fast with AI's Gen AI Launchpad - an 8-week hands-on program that takes you from zero to shipping real AI agents, apps, and automation workflows.
Register here: buildfastwithai.com/genai-course
References
1. Agent Studio: AI Agents by Razorpay (official launch post) - Razorpay Blog - March 12, 2026
2. Reimagining Merchant Experience with the Razorpay Agentic Platform - Razorpay Blog - March 12, 2026
3. Agentic Payments: The Future of In-App Commerce in 2026 - Razorpay Blog -March 12, 2026
4. Razorpay unveils world's first Agent Studio to automate payments - ANI News / The Tribune - March 12, 2026
5. Razorpay rolls out AI Agent Studio for payments - The Paypers - March 2026
6. Razorpay Launches World's First AI-Native Agent Studio - Enterprise IT World - March 2026
7. Razorpay Launches AI Agent Studio to Automate Business Payments - BankersAdda - March 2026
8. Building agents with the Claude Agent SDK - Anthropic / Claude -September 2025
9. Deep Dive: Agentic AI in Payments and Commerce- Fintech Wrap Up - June 2025
10. Razorpay CEO quote: 'A single person can operate like a team of 100 agents' - NewKerala / ANI - March 12, 2026


