Glean

Glean is an enterprise search and AI platform led by founder and CEO Arvind Jain. It expanded from search across workplace applications into conversational assistance, custom AI experiences, and agents. Its central technical foundation is company context: retrieving relevant documents, conversations, and relationships within a user’s access rights before a model answers a question or performs a task. That foundation began with the organization’s earlier work on workplace search.[1][3][12][17]

Glean Technologies, Inc. was founded in 2019 by Arvind Jain, T.R. Vishwanath, Tony Gentilcore, and Piyush Prahladka. In his September 2021 founder account, Jain traced the problem to his experience after co-founding Rubrik: information was scattered across applications, and an annual employee pulse survey identified difficulty finding documents and experts as the company’s biggest productivity challenge. Glean’s early product combined cross-application search, personalized results, and discovery of knowledgeable colleagues.[1][12][9][10]

In an April 2023 Latent Space interview, founding engineer Deedy Das recalled taking Google’s internal MoMA search tool for granted before leaving. The interview illustrated the workplace problem with a remembered presentation whose location was unknown: searching each application separately requires knowing where to start. It described richer SaaS APIs—including permission information—and distributed search infrastructure as conditions that made cross-application search practical. Das was a founding engineer, distinct from Glean’s four co-founders.[2][1]

In April 2023, Das described natural-language queries as a small fraction of Glean’s search traffic; project names, acronyms, and people were common queries. Glean also offered Go Links, a discovery feed, and consolidated Mentions across applications. Das described these employee-portal features as ways to keep employees returning beyond individual searches.[2]

Glean announced Glean Chat on June 6, 2023, adding conversational answering, analysis, summarization, and synthesis grounded in company information and user permissions. By February 2024, it also offered a low-code/no-code platform for building custom AI experiences using company knowledge.[4][12]

In How to build an AI assistant for the enterprise, dated July 2023, Chau Tran, Mrinal Mohit, and Arvind Jain explained the architecture: build a centralized index across applications, retrieve relevant information the user is authorized to access at query time, and pass it to a language model for synthesis. This is retrieval-augmented generation, or RAG—the model receives relevant evidence before composing its response. Permission filtering is part of selecting that evidence.[3]

Glean’s retrieval system combines keyword and semantic search with reranking, using freshness and personalization alongside topical relevance. Its knowledge graph represents content, people, activity, document links, and interactions, allowing relationships among users, teams, and authors to inform ranking. Exact project names and acronyms remain useful even when semantic matching helps find related material expressed in different words.[3][2]

The engineering account also describes company-specific embedding encoders, bootstrapped with synthetic examples to adapt retrieval to customer terminology; it says each customer’s data is used exclusively for that customer’s encoders. Query planning extends the approach beyond finding one document. In its example, a request about changes to Glean Chat becomes searches for GitHub pull requests and Slack-channel discussions before the results are synthesized.[3]

On May 20, 2025, Glean announced the general availability of Glean Agents. The offering included a natural-language builder, a library with more than 30 prebuilt quickstart agents, sharing, a reasoning engine, and a model hub with per-step model and temperature choices. Availability varied by component: the hosted MCP server, Agents API, deep research, triggering, looping, observability, and evaluations were beta in that announcement.[17]

Longer tasks introduced a memory-management problem. In their February 18, 2026 sandbox account, Ayushi Mrigen, Abhilash Samantapudi, and Matt Ding described context-trimming heuristics whose quality drifted as model versions changed. Their response was filesystem persistence: store additional task information outside the immediate model context and let the agent retrieve it later, with a command line, tool index, and Python execution for analysis. As they put it, “With the sandbox, memory management shifted from deletion to persistence.”[6]

The sandbox article’s sales-analysis example combines Salesforce opportunities with customer conversations and other unstructured enterprise information. Its design includes VPC isolation, scoped filesystems, session isolation, and network access disabled by default. The article described agent sandboxes as coming soon in February 2026.[6]

Glean’s August 26, 2026 agent update described proactive agents with provisioned access and their own identities. Its internal on-call example investigates escalations across Slack, Jira, and GitHub and drafts fixes. The update also described building agents through Claude Code or Cursor, Git synchronization, version checkpoints, collaborative editing, and Agent2Agent interoperability. Independent agents were beta; agent identity was public beta, built-in evaluations private beta, and agent scanning beta.[5][18]

The parallel August platform announcement extended Glean toward governed external AI access and model routing. Glean Intelligence, AI usage controls, and memory via MCP were generally available; the AI gateway, automatic routing, and team chat were beta. Glean Tau, a planned desktop workspace combining enterprise context with local files, applications, code, and models, remained coming soon.[18]

Glean also supplies company context to assistants outside its own interface. Its MCP documentation, updated September 11, 2026, describes retrieval tools, agents exposed as tools, custom and connector tools, and third-party MCP servers proxied through a gateway. It documents OAuth authentication, user-permission enforcement, and plug-ins for Cursor, Claude Code, and Codex, allowing external hosts to call Glean’s tools and agents.[7]

• Series C: a founder announcement labeled May 17, 2022 reported $100 million raised, led by Sequoia, at a $1 billion valuation.[11]
• Series D: on February 27, 2024, Glean announced over $200 million raised at a $2.2 billion valuation, co-led by Kleiner Perkins and Lightspeed Venture Partners.[12]
• Series E: a September 10, 2024 founder announcement reported over $260 million raised at a $4.6 billion valuation, co-led by Altimeter and DST Global.[13]
• Series F: on June 10, 2025, Glean announced $150 million raised at a $7.2 billion valuation, led by Wellington Management. Announced investment priorities included product innovation, partner ecosystem expansion, and international growth.[14]

Jain reported over 850 team members worldwide on June 10, 2025. On May 28, 2026, Glean announced it had reached $300 million in annual recurring revenue, 15 months after reaching $100 million ARR. ARR is a recurring-revenue run rate, distinct from recognized annual revenue. The company also reported nearly doubling its Fortune 500 customer count year over year.[15][16]