i2Coalition Member Spotlight: Gaffa
In 2023, James Mundy had been playing around with some ideas around web automation using LLMs: given a web page, could they navigate and perform tasks. We see how that turned out, and in 2024 Mundy founded Gaffa, a browser automation and web scraping API for developers and the agents they are now building with. We caught up with James to chat about simplicity, keeping robots in line, and the value of the open web.
i2Coalition: Can you give us the elevator pitch for Gaffa?
James Mundy: We make performing complex web automations as simple as sending a single request to our API: you send the actions you want to happen on the page and the format you want the data in and Gaffa handles the rest. We choose the right tools for the job under the hood and deal with all the complexities that retrieving public web data entails these days.
i2Coalition: How do you describe your browser automation as a service when you’re at a cocktail party?
JM: Web data is the lifeblood of the Internet economy that no one really acknowledges, yet most people use a product powered or enhanced by it every day. The data is public, but access isn’t. Larger companies invest huge amounts of dev time and resources keeping their pipes clear and data flowing whilst smaller companies are priced out of using the same data.
We want Gaffa to be a tool to help teams use web data effectively and reliably without having to have scraping experts on their team dealing with proxies, infrastructure, browsers and failures so they can focus on using the data to improve their product and customer experience.
i2Coalition: How has your customers’ use of AI influenced your service offerings?
JM: Initially we started off building primarily a developer product to be used by humans but it has become pretty apparent that AI will most likely be the user reading our docs, experimenting with our API and integrating it into customer solutions. To be honest, I’d say the experience of creating a product for agents and humans hasn’t been too different —a well-designed and documented API works best for both!
It’s been amazing to see how pairing Gaffa with Claude allows Claude to iterate on a site to create a reliable browser automation script without having to write any scraping code and even check the robots.txt and change code as a result to keep it compliant.
i2Coalition: What impact do you think recent geopolitical events have had on your more recent growth?
JM: At the moment we are a small business so I wouldn’t say geopolitics has had a massive effect on us, but obviously the global growth of AI and its capabilities has meant we can integrate features into the product that wouldn’t have been possible a year ago. As our business grows and the patchwork of global AI regulation increases, we will have to keep on top of it, something that being part of the Ethical Web Data Collection Initiative and i2Coalition will help with.
i2Coalition: Why did your team find it important to join the i2Coalition?
JM: As one of the smaller teams that have signed up to the EWDCI, instilling responsible and ethical practices from the start for us is important. I think the open web is a fantastic and rich resource that has been hugely beneficial to the world over the last 20+ years (mostly) and as laws are introduced to protect it, I want to ensure that we have a voice in the conversation which, as a small team, we otherwise wouldn’t have. Ensuring we have a framework of standards of how our customers use our products, who our suppliers are and how we verify our customers going forward will both be good for business and hopefully the wider Internet.
i2Coalition: Which i2Coalition initiatives connect most deeply with your team’s ethos?
JM: The EWDCI’s four pillars provide a simple, yet useful framework for thinking about how we grow our business, and how we educate customers on using the product. One of these, Social Responsibility, is a big reason why we are looking to establish free access for researchers and non-profits using web data for social good.
For example, we recently worked with a charity to help them reach the terms of service of social media and adult sites, so survivors of image-based online sexual abuse could write better take-down letters that cite platform policies.
i2Coalition: Where does government policy interface with your work on a daily basis?
JM: Day to day, it shows up mostly through our own acceptable use policy and who we let use the product, but looking ahead I’m particularly watching how rules written to regulate online data collection for AI training develop as they could affect even small businesses wanting to use online data if rules aren’t drafted to distinguish AI training from other uses of public web data.
