Beyond APIs: How AI Web Agents Are Automating the "Long Tail" of Knowledge Work
Arjun Chintapalli · Bhavani Kalisetty
AI Engineer Summit 2025 · 17:44
AI browser automation and web agent infrastructure
rtrvr.ai builds AI agents that carry out web tasks from natural-language instructions, including extracting datasets, filling forms and gathering sales leads. Its Retriever AI agent works across public websites and signed-in accounts through a Chrome extension, cloud execution, API, MCP, CLI and SDK. The product family also includes Rover, which runs demos, onboarding and setup inside software companies’ products, and Data & Evals, which supplies AI labs with browser trajectories, live environments, evaluations and verifiers.
Founded in 2024, the company is led by co-founder and CEO Arjun Chintapalli and co-founder and CTO Bhavani Kalisetty. Its origins lie in automating municipal website forms and retrieving audit files that search engines could not reach. The engineering approach uses Chrome extension APIs and structured representations of the live DOM, preserving page relationships and accessibility information. Code-as-plan converts model reasoning into sandboxed JavaScript, while AI Subroutines replay recorded tasks without model tokens during replay.
In 2026, the company reported more than 7 million tasks automated and adoption by more than 35,000 builders, go-to-market teams and founders. Retriever Auto’s extension agent is free and supported by sponsored cards. Its engineering work on prefix caching also reported a 90% reduction in browser-agent token costs, connecting operating costs to that advertising-supported model.
Arjun Chintapalli · Bhavani Kalisetty
AI Engineer Summit 2025 · 17:44
Affiliations reflect their AIE appearances, not necessarily current employment.
The presenters contrast their browser-native approach with cloud-hosted browser agents, introducing a deployment distinction alongside the task demonstrations.
Affiliations reflect each recorded session, not necessarily current employment.