AI Agents Break Out: TinyFish and Glean Bridge the Gap to the Live Web

PALO ALTO, CA – June 23, 2026 – For years, enterprise artificial intelligence has operated with one hand tied behind its back. While immensely powerful at analyzing a company’s internal universe of data, it has remained largely disconnected from the chaotic, ever-changing world of the live internet. A new partnership announced today between web infrastructure builder TinyFish and enterprise AI platform Glean aims to sever those restraints, empowering AI agents to move from passive insight to direct action across the open web.

The collaboration integrates TinyFish’s powerful suite of web-interaction tools directly into Glean’s secure AI environment. This allows AI agents that are already versed in a company’s private knowledge, permissions, and internal systems to now search public sources, fetch real-time data from live web pages, extract structured information, and execute complex, multi-step tasks across external websites and portals. It’s a move that signals a critical evolution in enterprise automation, bridging the long-standing divide between proprietary context and public information.

“Action is the next frontier for enterprise AI,” said Shuhao Zhang, Co-Founder of TinyFish. “Your agent should not just tell you the answer. It should complete the action for you. With Glean providing enterprise context and TinyFish providing live-web operation, enterprise agents can move from intent to completed work.”

From Secure Insight to Dynamic Action

Glean has established its position in the market by providing enterprises with a secure AI surface to reason over and act upon their vast stores of internal company knowledge. The platform gives AI a trusted foundation, respecting the intricate web of data permissions and internal workflows. However, much of the work that drives a business forward happens outside these digital walls—on procurement portals, competitor websites, regulatory filing systems, and logistics trackers.

This is where TinyFish provides the missing link. The company has developed a sophisticated infrastructure layer built on four core primitives for AI agents: a search API optimized for LLMs, a fetch API that cleans and structures live web content, a high-performance cloud browser for direct interaction, and a web agent capable of executing complex tasks from natural language commands.

The technical synergy is profound. A Glean agent, for instance, can now use TinyFish to monitor government procurement portals for new Request for Proposals (RFPs). Upon finding a relevant opportunity, it can fetch the live documents, extract key requirements and deadlines, and then cross-reference this information against its internal knowledge of the company’s capabilities, past bids, and strategic criteria. The result is a nearly complete response package, prepared autonomously but leaving the final strategic review and submission in human hands.

“Enterprise AI becomes more valuable when it can connect trusted company context with the outside world in real time at scale,” noted Zubin Arani, VP of Partnerships at Glean. “TinyFish extends that capability into live web environments, where agents can navigate dynamic pages and workflows to help people move work forward.”

The Engine Room of Web Automation

The power of this partnership lies in the robustness of TinyFish’s underlying technology, which was engineered specifically to overcome the fragility of traditional web automation. The platform is already handling millions of workflows a month for enterprise clients, built on a forked Chromium engine that moves automation into the native layer. This architecture allows it to bypass many common anti-bot defenses and deliver remarkable performance.

Research confirms impressive metrics: search responses average under 600 milliseconds, and new cloud browser sessions launch in under 100 milliseconds—a stark contrast to the 5-10 second cold starts common with other managed browser services. This speed enables massive parallel execution. One documented test showed the platform could scan 50 different supplier portals, a task that would take a human over 45 minutes, in just over two minutes.

Crucially, TinyFish’s fetch and browser layers are designed for AI consumption. Instead of feeding a language model a messy dump of raw HTML, the platform extracts and returns only the clean, relevant content in a structured format like Markdown or JSON. This can reduce the number of tokens an AI agent needs to process by up to 90%, dramatically lowering operational costs and improving processing speed.

Furthermore, its web agent uses Natural Language Control, interpreting human-like instructions to navigate sites and complete forms. This makes automations far more resilient to the constant layout and code changes that plague traditional RPA bots, which often rely on brittle CSS selectors that break with the slightest website update.

Redefining the Automation Landscape

The combined TinyFish-Glean offering enters a competitive field, but its integrated approach presents a clear differentiator. It challenges not only traditional RPA players like UiPath but also the fragmented ecosystem of individual web scraping tools, proxy services, and browser APIs that developers currently stitch together. By unifying search, data fetching, browser control, and agentic action under a single, reliable infrastructure, the partnership offers a cohesive solution for what is being termed “Agentic Automation.”

The applications extend far beyond RFP preparation. In logistics, agents can monitor real-time stock levels across dozens of supplier websites. In finance, they can track competitors’ pricing changes on authenticated portals. For market intelligence teams, they can automate the collection of product data, sentiment analysis from forums, and financial details from corporate filings. According to one industry expert, this ability to reliably automate workflows on authenticated, JavaScript-heavy sites is a capability that many have found elusive until now.

The future of work is increasingly intertwined with such intelligent agents acting as digital assistants. This partnership provides a blueprint for how these agents can be safely and effectively deployed, combining the secure, permission-aware context of the enterprise with the dynamic, real-time reality of the open web. As businesses continue to seek efficiency gains, the ability for AI to not just analyze but to act will become the new standard for progress.