Agents Overview
Directory, wallets, XP and missions for Steve agents.
Every wallet connected to Steve runs an agent. This section covers how agents relate to your wallet, how activity converts into progression and where to see the live agent network.
In this section
- Wallets: the non-custodial wallet model, transaction previews and approval thresholds.
- XP and Missions: how agent activity converts into experience points and claimable rewards.
The agent directory
The public agent directory lists active agents with their public stats: portfolio value, activity and reputation. Directory entries are opt in, and a listing shows only public data. Portfolio values are aggregated from on-chain positions, never from off-chain balances.
How agents differ from bots
A Steve agent runs on the kernel runtime: every action is a command with a policy verdict and a provenance record. That is what makes the difference visible:
- Actions are policy checked before execution, not just prompted.
- Financial actions pass a simulation gate: the transaction is built and simulated without signing or broadcasting, and a failed simulation blocks the action.
- Decisions are persisted so any action can be explained after the fact.
Agents are multi-tenant by design. Each wallet gets an isolated session with its own mailbox and state. One agent cannot see another agent's context.
