MCG Live · November 11, 2025 · 42 min
Foundry's Unbrowse is an AI agent infrastructure that learns how websites work at the packet level rather than through UI clicking, enabling agents to access any web service via API calls. Founder Louis claims this approach is faster and more reliable than traditional computer use agents, with a token model that rewards indexers, websites, and token holders. The project is in beta with ~20 testers and aims to build a crowdsourced index of web abilities for AI agents.
Unbrowse operates at the network packet layer instead of UI level, allowing agents to call APIs directly rather than clicking through web interfaces
Indexing is crowdsourced: users who browse websites teach the AI how to interact with them at the packet level, creating collective intelligence
Unbrowse integrates with MCP (Model Context Protocol), working with Claude, ChatGPT, or any AI agents to provide web access without code changes
Four-way payment split when agents use indexed abilities: indexer, website, investors (Foundry holders), and infrastructure each receive a portion, with investor portion burned or staked
Demo shows Unbrowse pulling inventory data in ~3 seconds versus 28 seconds for UI-based browser agents like Claude or Ellis
Foundry received Nvidia inception program backing with cloud credits from Google Cloud and AWS to handle compute costs
Currently in beta with approximately 20 testers; founder keeping user base limited to maintain first-mover advantage on the indexed abilities flywheel
Use case: agents can analyze web data like Google Trends and earnings reports to predict stock performance, as founder previously did manually
This summary was written from the recording's transcript by a language model. For the exact words, ask the MCG Live archive: every answer there is quoted word for word and linked to the second it was said. Market Bubble Search is independent and not affiliated with the show. Nothing here is advice.