How the agent confirms the intention
While the agent operation is the dialogue with the agent to make a decision, the MetaMask popup is the conclusion of this conversation.
Think of it as the digital equivalent of your financial advisor explaining a strategy and then handing you the final document to sign. Signing is deliberate and necessary confirmation that you understand and consent to the action. This can turn the agent’s recommendation into reality with your explicit approval, providing crucial peace of mind. Especially since an agent’s interpretation can vary widely depending on the underlying LLM, conversational context, and the data they have access to, it’s always a good idea to double-check a portfolio transaction before approving it.
MCP servers should serve both realities
The vision of a future in which agents autonomously pay other agents for services requires the agent-controlled model. The agents of this economy will need their own capital to function.
However, the transaction creation model provides a secure gateway to this future. It can be used to securely fund an agent or simply to execute one-off transactions for simpler operations. This flexibility is essential.
From a developer’s perspective, adding this feature shouldn’t be a big task. If an MCP server can already prepare and sign a transaction with a key it holds, it should be able to perform the same logic without the final signing step, instead returning the transaction unsigned. This minor change paves the way for a much more secure and flexible paradigm for users and may even enable more complex designs, such as a dedicated “signing agent” in a multi-agent system.
Therefore, any robust MCP server designed for widespread adoption should provide developers with the flexibility to create applications capable of:
-
Advise and develop secure, user-centered financial decisions.
-
Run with delegated funds for specialized, automated, clearly defined tasks.
We recommend continuing this dual support to promote real innovation while protecting users.
You can learn more about create Web3 agents using Google Cloud here.
