Indian startups building in the AI agent space are overwhelmingly targeting the orchestration and application layers rather than competing with OpenAI, Anthropic, or Google at the foundational model level. Inc42 reports that following NVIDIA’s GTC 2026 event and the launch of NemoClaw (NVIDIA’s enterprise wrapper around OpenClaw), the race to own the “control layer” of the agent stack is intensifying, with Indian companies positioning themselves as the middleware between frontier models and enterprise workflows.
Razorpay: Agent Studio on Anthropic’s Claude SDK
The highest-profile example is Razorpay, which launched its Agent Studio and Agentic Experience Platform at its annual FTX 2026 event in Bengaluru on March 12. Built on Anthropic’s Claude Agent SDK, the platform lets businesses deploy AI agents that handle payment operations: abandoned cart recovery, dispute responses, subscription recovery, and cashflow forecasting.
Razorpay’s Chief Product Officer Khilan Haria told Inc42 that the company deliberately avoids building foundational models. “That’s Anthropic’s and OpenAI’s domain,” Haria said. “Instead, we focus on making that intelligence actionable within real-world commerce contexts, at scale.”
The platform is being tested with Swiggy, Zomato (via its Nugget product), PVR Inox, BigBasket, and LinkedIn as partners. Businesses can describe tasks in plain English and connect agents to external systems including Shopify, WhatsApp, Slack, Tally, and QuickBooks, according to CIOL’s coverage of the launch.
Voice Agents: Gnani.ai and Bolna AI
Bengaluru-based Gnani.ai launched Inya, a multi-agent platform for building and deploying enterprise voice agents. CEO Ganesh Gopalan told Inc42 that the platform includes prebuilt workflows and an orchestration layer that manages interactions at scale while minimizing latency. “In many cases, partners and customers have been able to build and deploy voice AI agents within 30 minutes,” Gopalan said.
Bolna AI, also based in Bengaluru, operates in the orchestration layer for multilingual voice agents across different call scenarios. Its pricing starts at ₹2.5 per minute (roughly $0.03 USD), targeting the cost sensitivity of the Indian market.
Noida-based Squadstack focuses on the application layer, building orchestration capabilities specifically for revenue and customer experience workflows in production environments.
The Middleware Thesis
The strategic logic is straightforward: as foundational models commoditize and inference costs fall, the companies that control how agents are orchestrated, integrated, and monetized will capture disproportionate value. Ashvin Vellody, Partner at Deloitte India, told Inc42 that companies now recognize “value will not come from the model alone. It will come from making the technology easier to use for a much larger pool of developers and builders who can create product-grade solutions.”
Arun Chandrasekaran, VP analyst at Gartner, reinforced the point. “While frameworks can be monetised, the real monetisation will likely be indirect,” Chandrasekaran told Inc42. “Revenue will ultimately come through models, infrastructure, and applications.” For India specifically, Chandrasekaran noted that lightweight, modular, and open frameworks are likely to find greater traction than heavy enterprise stacks.
Where This Fits in the Broader Stack
The Indian orchestration bet sits in a specific layer of a rapidly crystallizing agent infrastructure stack. At the bottom, NVIDIA provides compute and NemoClaw for enterprise agent deployment. In the middle, platforms like Razorpay’s Agent Studio and Gnani.ai’s Inya provide the workflow orchestration connecting frontier models to business logic. At the top, purpose-built agents execute domain-specific tasks.
The major Western players are building vertically across multiple layers: Microsoft with Agent 365, Google with Vertex AI Agent Builder, Salesforce with Agentforce. Indian startups are betting that horizontal specialization at the orchestration layer, combined with lower price points and multilingual capabilities, creates a defensible position even as the bigger players consolidate above and below them.