Definition
An agent wallet is a programmable wallet or account that lets a software agent request or initiate transactions within permissions set by a person or application.
How an agent wallet works
The person does not need to give the agent the master wallet credential. Instead, the product can issue a narrower credential or delegated signer. When the agent asks to act, the wallet, account or service checks the request against the permissions attached to that credential before a transaction is signed or submitted.
Those permissions may cover order size, approved assets, approved venues, expiry and revocation. Their presence is not implied by the phrase agent wallet. They only exist when the product implements and enforces them.
What to check before using one
Ask who controls the main key, where funds are held, where limits are enforced, how access expires and how the agent can be stopped. An agent wallet can be user controlled, venue held or split across several systems. The name describes what the wallet is used for, not its custody model.
What it does not mean
An agent wallet is not automatically self-custodial, and it does not automatically make an agent safe. Custody and control depend on the product design.
Where this term fits
Questions about agent wallet
- Does an AI agent own an agent wallet?
- No. The agent is software acting through a permission or credential. Who owns or controls the wallet depends on the custody and key design of the product.
- Is an agent wallet the same as a normal crypto wallet?
- It can use the same blockchain rails, but it adds a way for software to request actions without receiving unrestricted access to the main wallet credential.
- Can an agent wallet enforce spending limits?
- Some can. Check whether the limit is enforced by the wallet or account, or merely written as an instruction for the agent to follow.
See these concepts in context
Review the Fensory MCP connection, tools and account controls.