AI agents are quickly emerging as the next major interface for users—on par with, and perhaps even replacing social feeds. Instead of scrolling, users will interact with high context human like counterparts that anticipate needs, retrieve insights, and personalize every experience—from content to transaction recommendations.
At Embed, we’ve been collaborating with teams building in the AI agent space to explore how semantic search and personalization can power this next generation of agentic interfaces. One of the most forward-thinking builders we’ve worked with is Zo. The Zo team is building tools to make agents not just more capable—but deeply insightful and personalized.
" You're going to stick with the agent that knows you the best, that understands you, and that's going to be very sticky.” Azi, Zo Founder & CEO
In this piece, we’ll walk through how Zo is using the Embed semantic search API to give their Farcaster discovery agent superpowers.
Farcaster open data + AI = smarter, more personal agents
When a user opens the Zo app—similar to a chat interface like Telegram or WhatsApp—they’re assigned a default generalist agent that’s able to provide information on things as diverse as ERC20 contracts and the weather. But most interestingly, the Zo marketplace has a selection of access mini apps tailored to them.
One of the most compelling examples so far? The Farcaster Discovery Agent, powered by the ’mbd semantic search API.
This agent offers a glimpse into the future of how users will engage with crypto-native social content: through personalized, conversational interfaces.
On the image below is a search query about a niche NFT collection in the $higher token community on Farcaster:
Sample query about a trending NFT collection on Farcaster, with a high-quality synthesis of recent activity.
Using the Farcaster Discover bot is simple:
Open the Zo app and search for Farcaster Discovery, or use this direct link.
Ask the agent about communities, events, or projects trending on Farcaster. The better the prompt, the richer the result—just like a good RAG (retrieval-augmented generation) system.
The agent takes your query, builds a semantic search request via the ’mbd API, and returns a clean, synthesized answer based on relevant onchain social content. See the documentation for our search API here.
Over time, the Zo team aims to make these agents even smarter—connecting them to a broader personalization engine powered by social graphs, wallet history, and topic interest.
Why this matters
AI agents are evolving quickly—but the winning ones won’t just be more capable. They’ll be more context-aware, more personal, and more social.
At ’mbd, we’re helping builders create agents that tap into web3’s unique data layers—onchain transactions, decentralized social graphs, and real-time content—to make agents more useful. Whether it’s through semantic search, topic-specific content feeds, or personalization APIs, we’re enabling agents to deliver the right insight at the right time.
According to Zo’s founder Azi, the next generation of agents will monetize not by showing content, but by executing on intent—taking small cuts on actions they help initiate, from swaps to purchases. In this world, attention, retention, and personalization become the battleground. Agentic apps are now competing for engagement—and the smartest ones are those that truly know their users.