Wordware is a central player in the shift from "AI as a chatbot" to "AI as an agentic workspace." They are active at the application layer of the agent stack, specifically focusing on the orchestration and memory components. Their pivot from a developer IDE to the Sauna workspace reflects a broader industry trend where the complexity of agent construction is being abstracted away in favor of seamless, demonstration-based automation.
For those building or using agents, Wordware is a case study in how to handle long-term context and human-in-the-loop validation. They are championing a model where agents are not just reactive but proactive—running batch operations in the background while the user sleeps. By focusing on "visual trust" and "programming through demonstration," they are pushing forward the standards for how humans collaborate with autonomous agents without losing control over the final outcomes.
Wordware is an AI context lab based in San Francisco, currently operating from an office at 1185B Old Mason Street. Founded in 2023 by Filip Kozera and Robert Gaal—who met a decade prior while studying deep learning at the University of Cambridge—the company is perhaps best known for raising $30 million in seed funding led by Spark Capital and Felicis. At the time, this was the largest seed round in Y Combinator history. The firm’s early thesis centered on the idea that "words are the next programming language," leading to the creation of a popular IDE for prompt engineering and agent development.
In mid-2025, the company underwent a significant pivot. Despite having thousands of developers using their infrastructure, the founders observed that the most successful implementations weren't developer tools but assistants that handled work naturally. They realized that for most users, the act of building a workflow was an unnecessary tax. Users did not want to build automation; they wanted their work done. This realization shifted Wordware’s focus from providing the tools to build agents to building the definitive agentic workspace: Sauna.
Sauna is designed as an intelligent workspace where context is the foundation rather than a feature. Most current AI interactions are limited by sliding context windows that eventually forget previous details. Wordware’s architecture instead uses a queryable file system that stores a user’s emails, transcripts, documents, and uploaded files. This persistent memory allows the AI to learn a user’s specific taste, preferences, and decision-making patterns over time.
This system enables what the company calls "programming through demonstration." When a user performs a task—such as evaluating a candidate or drafting a memo—Sauna observes. By the third repetition, the system detects the pattern and offers to automate the process proactively. The goal is to move from manual assistance to background execution, where the agent handles repetitive "grunt work" while the human retains final judgment and approval power. This human-in-the-loop model ensures accountability while maximizing the leverage of small teams.
Wordware is explicit about its competitive stance: it cannot out-train foundation model providers like OpenAI or Google on raw intelligence, so it competes on product. They argue that as intelligence becomes a commodity, the winning companies will be those with the deepest access to user context. By integrating with over 2,000 tools through their "Triggers & Actions" system, Sauna aims to become the central nervous system for knowledge work.
The company is currently a team of 14 people, maintaining a lean headcount even after its massive funding. They are moving their infrastructure to the cloud in late 2025, with a broader public launch for Sauna planned for early 2026. The bet is that the next generation of AGI will not be a smarter model, but a product that millions of people trust enough to integrate into every part of their digital lives.
An intelligent workspace that eliminates busywork and anticipates needs through persistent memory.
Wordware is hiring.