Bengaluru: Composio, a San Francisco-based infrastructure startup, has raised $25 million with its latest funding as it aims to build basic tools for creating artificial intelligence (AI) agents that can be learned through experience. The integrated startup that simplifies how AI agents and large-scale language models (LLMs) connect with external applications and services was founded by Indian entrepreneurs Soham Ganatra and Karan Vaidya.The latest $25 million Series A round will be led by Lightspeed Venture Partners and institutional investors including Vercel CEO Guillermo Rauch, Hubspot founder Dharmesh Shah, Marathon Management Partners founder Gokul Rajaram, Rubrik founder Soham Mazumdar, SV Angel, Blitzscaling Ventures and Operator Partners Fund The fund also took part in the round along with existing backer's elevation capital. The startup had previously raised $4 million in seed funding.Composio builds what we describe as a shared learning layer for AI agents, allowing us to accumulate and transfer practical knowledge throughout our workflow. For example, when an agent learns how to handle Salesforce or GitHub Edge cases, that insight becomes instantly available across the network, making the agent more useful over time rather than staying static.“I could spend hundreds of hours building LLM tools, adjusting prompts, and polishing instructions, but I hit a wall,” said Ganatra, CEO of Composio. “These models don't get better at work like human employees do. We're solving this on the infrastructure level.”The company began building stacks in 2022 and has since tackled challenges related to multi-agent tuning, secure authentication and scalable architecture. The core components of the platform are reinforced learning layers designed to help AI develop intuition.Composio's tools bring together over 100,000 developers and over 200 companies, including several Y-combinator startups such as Glean, April, OpenNote, Airweave, Den and Dash.With the new funding, Composio stated its plans to deepen its learning infrastructure and expand its integration with AI frameworks such as Supabase MCP, Langchain, Vercel AI SDK and Openai Agent.“Composio is building a shortage of layers where AI agents really help production,” said Raviraj Jain, partner at LightSpeed. “By enabling agents to learn from experience, they bridge the gap between impressive demonstrations and real-world developments.”
