Fetch.ai has published a new developer walkthrough showing how specialized AI agents can connect with ASI, including agents capable of handling image-based tasks. The example extends ASI beyond a standalone conversational model by showing how it can discover and communicate with independently deployed agents that provide specialized capabilities. Fetch.ai highlighted the workflow in an official announcement, positioning multimodal agents as another building block in its wider agent ecosystem.
The implementation relies on Fetch.ai’s uAgents framework, Agentverse and the standardized Agent Chat Protocol. The project’s official ASI agent documentation shows developers how to create an agent, connect it through an Agentverse Mailbox and publish its protocol manifest for discovery. The key requirement is interoperability: an agent must expose its capabilities through supported protocols before ASI can reliably identify and engage it.
Want your AI Agent to understand images? Build an Image Analysis Agent in 7 steps 👇
1. Create a hosted Agent on https://t.co/MXpAx3L9P4
Start with a new hosted Agent and add:
'https://t.co/xpU8xjkABD
image_analysis.py'2. Add the Agent Chat Protocol
Handle:
ChatMessage… pic.twitter.com/gYujlGaStK
— Fetch.ai (@Fetch_ai) September 21, 2026
Agentverse Provides the Discovery Layer
Under the architecture, uAgents handles agent logic and communication while Agentverse provides hosting, profiles and discovery infrastructure. Agents register through the Almanac, a directory on the ASI Network that stores their addresses and supported protocol information. Once properly registered, an agent becomes accessible across the ASI ecosystem and can be located by ASI according to the capabilities it advertises. Registration and published metadata are therefore part of the discovery process rather than an automatic consequence of writing the agent code.
For a multimodal use case, the agent can be designed to accept image-related inputs, route them to a compatible vision model and return structured results through the same communication framework used by other agents. The image-analysis functionality comes from the model and agent logic, while Agent Chat Protocol standardizes how that capability is exposed to ASI and other agents. That separation allows developers to change the underlying AI model without rebuilding the broader discovery and communication layer.
Fetch.ai has been expanding this interoperability model through tooling for single-agent and multi-agent architectures. Current uAgents documentation also includes adapters for systems such as CrewAI, LangGraph and external agent frameworks, allowing existing AI workflows to be bridged into Agentverse. The broader strategy is to make ASI an orchestration layer that can call specialized agents instead of requiring one model to perform every task directly.
Deployment Does Not Guarantee Agent Usage
Agentverse currently advertises an open directory containing roughly 2.7 million agents, but registration alone does not guarantee that a newly deployed agent will receive meaningful traffic. Fetch.ai’s own tooling emphasizes metadata quality, capability descriptions and discoverability because ASI must determine which agent best matches a user request. The practical challenge therefore shifts from simply deploying an agent quickly to making its function clear enough for other agents and users to find and use it.
That distinction also applies to Fetch.ai’s broader claims that developers can deploy agents rapidly. Fast deployment can reduce technical friction, but it does not by itself demonstrate accuracy, reliable multimodal performance or sustained demand for the resulting agent. Image-analysis quality will still depend on the selected vision model, task design, prompting, latency and the execution environment supporting the agent.
The next useful milestone will be measurable use of these specialized agents through ASI rather than another deployment tutorial alone. Discovery rates, completed agent-to-agent requests and real application integrations will show whether Fetch.ai’s architecture is evolving from developer tooling into a functioning market for specialized AI services. For now, the new walkthrough provides a concrete path for making multimodal agents interoperable with ASI while leaving model quality and adoption as separate questions.
Natalie Pierce tracks the parts of crypto that move fast and rarely wait for everyone to catch up. From South Africa, she covers DeFi, AI crypto, hacks, airdrops, sentiment and emerging narratives, especially when user behavior and protocol risk start to overlap.
Her reporting is built for messy sectors. Natalie looks at incentives, reactions, security concerns, social momentum and early signs of traction without pretending every new trend is already proven. Her voice is clear and accessible, but careful enough for areas where excitement can outrun the facts very quickly.