Composite Secures $5.6M Seed Funding for Cross-Browser AI Agent Tool

Composite gets backing from NFDG for its cross-browser agent tool | TechCrunch

The central pitch of AI browsers like Perplexity's Comet, Opera's Neon, and The Browser Company's Dia is that they'll help you complete daily tasks more efficiently. While these are agents that are limited to one browser, Composite aims to build an agentic solution that helps professionals with their tasks, regardless of which browser they use. The startup was started earlier this year by Yang Fan Yun and Charlie Deane. Yun is a former product manager at Uber, while Deane founded a company selling server proxies. When he was at Uber, he realized that a lot of people around him were doing grunt work in their browsers. "I saw people in different roles, including marketing, sales, recruitment, and security engineers, do a lot of tedious work in their browsers. I felt that this stops them from putting their education and skill to full use, and I wanted to automate this work easily for them," Yun told TechCrunch over a call about Composite's core problem statement. The company said today that it has raised $5.6 million in seed funding in a round led by Nat Friedman and Daniel Gross's venture firm NFDG, with participation from Menlo Ventures and Anthropic's Anthology Fund. Composite offers its solution for Macs at the moment, and setup is very easy, as you just need to install the browser extension for agents to make use of that browser. You can issue different commands across tools that you use on the web, and Composite will get the work done. For instance, it can help you go through your Jira backlog for bugs using relevant documents, leave comments on high-priority bugs, and also mark duplicate bugs as resolved. The startup says security engineers can use it for looking up candidates across sites and draft personalized emails, security engineers can create vulnerability tickets based on alerts, and marketers can pull reports from different sources to create short insights reports. Yun said that other AI browsers and agents from companies like OpenAI and Perplexity are solving for non-professional needs, such as helping with shopping and booking tickets. "We are an ideal tool for professionals who want to set their workflows without having technical knowledge. Composite is very good at atomic actions like clicking on different elements of the website or typing in boxes, and that gets the job done for our users," he said. He added that because the agents work in browsers where you are already logged on to services, it doesn't need connectors and can work across different sites. The tool already suggests some tasks based on users' patterns. But in the coming months, the company is aiming to develop a better mechanism to automatically surface tasks that Composite can do on the user's behalf. The startup is also working on a way to schedule tasks for recurring usage. The startup said that because it doesn't need users to switch browsers, allows admins to restrict tools, execute tasks locally, and also lets users define what websites are out of bounds, its tools are better suited for professionals. There is a lot of competition in the area of having agents do work for professionals. Companies like OpenAI that use their own browser to take actions, Notion relies on users' context within the app, combined with other connectors. General Catalyst-backed Highlight wants to use your whole desktop as context. A bunch of startups are working on a narrower context, like a spreadsheet. A lot of these startups are in early stages, and there are questions around AI agents' efficiency in the long term. While investors are ready to put their money, the startups working in this area will have a lot to prove to justify it. Matt Kraning, a partner at Menlo Ventures, is confident in Composite's ability to stand out. He told TechCrunch over a call that the tool was very intuitive to use for professionals without being overtly technical. "Composite handles different modalities and sites very well, and it is designed with professional use cases in mind. The tool is well-suited for people who might have to go through a lot of tasks in a day across a range. of functions," he said.

This browser-based AI wants to kill the worst part of your job

Composite, a startup that turns web browsers into intelligent automation engines, has raised $5.6 million in seed funding to eliminate what its founders call the "digital grunt work" plaguing millions of knowledge workers. The round was led by Nat Friedman and Daniel Gross of NFDG, with participation from Menlo Ventures and Anthropic's Anthology Fund. The San Francisco-based company has attracted users from hundreds of companies including Google, Uber, DoorDash, Tesla, Salesforce, and Reddit in just two months since launching. Unlike the wave of new AI browsers from companies like Perplexity and OpenAI that require users to abandon their existing setups, Composite takes a different approach: it transforms whatever browser professionals already use into an intelligent assistant that predicts and automates repetitive tasks. "We're not trying to replace Chrome," said Yang Fan Yun, Composite's co-founder and CEO, who was New Zealand's national valedictorian and achieved the highest GPA in Stanford's Computer Science department. "We're making the browser you already use infinitely more powerful by predicting and automating your work before you even think to ask." AI browser wars intensify as OpenAI, Google, and startups battle for automation dominance The funding comes as the AI agent space experiences explosive growth, with major players racing to build systems that can autonomously handle digital tasks. OpenAI recently demonstrated its "ChatGPT Agent" capabilities, while Perplexity launched its Comet browser, and Google continues developing AI features for Chrome. But Composite's investors argue the company has found a more practical path to adoption by meeting users where they already work rather than forcing them to learn new tools. "Quality and habit formation," said Matt Kraning, partner at Menlo Ventures, when asked what differentiates Composite from competitors. "They are meeting users where they are currently and not asking them to change huge numbers of habits at once, say, by having them switch to an entirely new browser." Kraning, who has built billion-dollar category creators, sees particular value in what he calls "messy middle" automations -- tasks that need to be done more than 2-3 times but fewer than tens of thousands of times. "I use it very frequently to help research companies and founders," he said. Inside Composite's technology: How local AI agents learn your work patterns Composite operates through a Chrome extension that monitors user behavior to learn individual workflows, then suggests and executes automations across any website. Users activate it with a keyboard shortcut that brings up a lightweight overlay showing personalized task suggestions. The system runs entirely locally on users' devices, addressing privacy concerns that have slowed enterprise AI adoption. "We never actually have to ask you for your login information or your credit card details," Yang explained, noting that the system leverages credentials already stored in users' browsers. The company uses multiple AI models under the hood -- combining small, fast open-source models for certain tasks with larger vision models for complex operations -- rather than being locked into a single provider's ecosystem. "We use best-in-class models from multiple providers, unlike the model labs who are limited to their own systems," Yang said during a recent interview with VentureBeat. "We also build our own proprietary technology to integrate everything seamlessly." From Uber to Tesla: How professionals save hours weekly with browser automation Early customers report significant productivity gains from automating previously manual processes. Kailiang Fu, a Product Manager at Uber, said Composite "lifts from my shoulders what used to take me hours of work per week, such as updating project trackers and running data queries." Yang described one security engineer at a major tech company who uses Composite to automate weekly security architecture reviews -- a process that previously required manually gathering information from GitHub, Confluence, Google Drive, and internal dashboards before synthesizing it into reports. "This is a really, really tedious process," Yang explained, "because it uses a lot of disparate internal tools that don't have APIs." The engineer now uses a single prompt to have Composite research across all these systems and generate the required documentation. Balancing AI surveillance with enterprise security: Composite's privacy-first approach The prospect of AI systems monitoring all browser activity raises significant privacy and security concerns, particularly for enterprise customers handling sensitive data. Composite has built several safeguards to address these issues, including blocklists for sensitive websites, explicit user confirmation for high-risk actions, and opt-out options for data collection. "We take data security, privacy and safety all really seriously," Yang emphasized, noting that Anthropic's investment reflects shared values around AI alignment and safety. The system maintains user control by running in visible browser tabs rather than remote servers, allowing workers to easily stop or override automated actions. "It's extremely low friction for you to close it, for you to stop it, for you to interact at any given time, because it's in your browser," Yang said. Viral growth and enterprise adoption: Why Composite gained thousands of users in two months Composite's rapid growth -- reaching thousands of users across hundreds of companies in just two months -- suggests strong market demand for browser automation tools. The company reports primarily organic growth through word-of-mouth adoption within organizations. "We have a user at a very large tech company who's gotten their whole team to use it, and it's spread to different teams across the organization," Yang said. This viral adoption pattern has generated enterprise inquiries from managers seeing multiple team members use the tool. The timing appears favorable as organizations increasingly seek AI solutions that integrate with existing workflows rather than requiring wholesale technology changes. According to Gartner, more than half of employees report spending over 50% of their working day on repetitive tasks -- copying data between systems, managing email, and updating project trackers, while 25% claim these tasks occupy over 75% of their workday. Why Composite believes it can beat Google and Microsoft at their own game While tech giants like Google, Microsoft, and OpenAI develop their own browser AI capabilities, Composite's investors believe the startup's focused approach provides advantages. "You have large incumbents that don't want to do the best thing for agents, because they need to protect their ads, and they need eyeballs to look at their ads manually," Yang argued. "Neither of them really are focused on what we're trying to do, which is just being maniacally focused on being the best predictor and executor." Kraning from Menlo Ventures sees the competition differently: "Don't want to take too much thunder away from what Yang and Charlie will be launching, but we're not going to be selling a tool, we're going to be selling outcomes and a system that changes how people and organizations do work." Technical hurdles and scaling challenges in long-running AI automation tasks Despite early success, Composite faces technical challenges around long-running, complex tasks. Yang acknowledged that automating workflows requiring thousands of sequential steps -- like processing hundreds of support tickets -- remains difficult due to AI model context limits and the need for parallel processing. "How do we parallelize this? How do we run these 400 processes at once and then bring it back together? How do we maintain planning throughout all of them?" Yang said, describing current engineering priorities. The $5.6 million funding will accelerate product development and go-to-market efforts, with plans to enhance both single-user features and add enterprise collaboration capabilities. The future of work: From digital grunt work to human potential Yang's ultimate vision extends beyond productivity tools to fundamentally changing how people experience work. "Most folks spend their working hours in front of a desk doing fairly mind-numbing work, watching the clock until 5pm so they can go home and do things they actually enjoy," he said. "We want to help people get into flow state and actually achieve their human potential. I think that's a really powerful mission." The company plans to expand Windows support and launch new personalization features that proactively identify and automate repetitive tasks without explicit user requests. Yang envisions a system that knows users so well it can predict their needs: "It's like a perfect boyfriend -- you need to know exactly what you want before you even tell us." As enterprise interest builds and tech giants mobilize their resources, Composite faces the classic startup challenge: scaling fast enough to establish market position before larger competitors can replicate their approach. The company's bet on augmenting existing browsers rather than replacing them offers a pragmatic path to adoption, but also leaves them vulnerable to platform changes by Google, Microsoft, and Apple. The stakes extend beyond Composite's commercial success. If the company's vision proves correct, the very nature of knowledge work could shift dramatically -- from humans drowning in digital busywork to AI handling the mundane while people focus on creativity, strategy, and genuine problem-solving. For millions of professionals checking email at 4:55 PM and dreaming of more meaningful work, that future can't arrive soon enough.