> ## Documentation Index
> Fetch the complete documentation index at: https://docs.privy.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Agent wallets

Privy enables you to provision wallets for AI agents.

Agent wallets are designed for builders shipping agents that need to pay for APIs, move funds between accounts, execute onchain transactions, or interact with DeFi protocols. Privy provides the key management, authorization model, and policy controls to let agents transact safely within human-defined boundaries.

Read below to learn how Privy powers agent wallets for autonomous trading systems, AI-powered payment pipelines, API monetization infrastructure, and general-purpose agentic applications.

## Authorization models

Privy supports two authorization models depending on whether an agent controls its own wallet or acts on behalf of a user.

**Agent-owned wallets** are wallets your agent provisions and controls. The agent holds the credential needed to trigger signing and can transact independently without prompting a user. This model suits fully autonomous agents that manage their own funds and operate outside any user session.

**Delegated signing** lets an agent transact on behalf of a user's existing wallet. Users grant the agent an [authorization key](/controls/authorization-keys/owners/overview) with a defined scope via policies. the agent signs transactions within that scope, and the user can revoke access at any time. This model suits agents that act as a user's financial representative while the user retains ultimate custody.

## Policies

Privy's [policy engine](/controls/policies/overview) allow you to enforce guardrails on how agents spend funds.Policies can restrict which tokens and chains an agent can transact on, cap transfer values per transaction or over a time window, allowlist recipient addresses and smart contracts, and more.

## Agent payments

Privy supports both [MPP](/recipes/agent-integrations/mpp) and [x402](/recipes/agent-integrations/x402). These are HTTP-native payment protocols that let agents pay for APIs, data feeds, and digital content autonomously without a manual checkout flow.

Privy's x402 and MPP clients intercept the 402 HTTP responses, sign payment authorizations with an agent's wallet, and retry the request automatically. This makes autonomous API monetization a drop-in behavior rather than a separate payment integration.

Both protocols work with Privy server wallets and are supported across React (via `useX402Fetch`) and Node.js (via `createX402Client`), with per-request `maxValue` caps to prevent runaway spending.

## Tooling and framework integrations

The [Privy Agent CLI](/recipes/agent-integrations/agent-cli) lets agents spin up, fund, and transact with wallets directly from the command line — a natural fit since agents already execute shell commands. The CLI uses a device authorization flow so a human approves wallet access in a browser once; subsequent transactions use short-lived ephemeral signing keys without ever exposing the private key to the agent process. Sessions last up to 30 days with automatic token rotation.

Privy's wallet infrastructure is framework-agnostic and integrates with LangChain, Vercel AI SDK, OpenAI Agents SDK, AWS Bedrock AgentCore, and custom agent runtimes. For MCP-based agents, Privy wallets can be connected to [MCP servers](/recipes/agent-integrations/overview) to expose signing as a tool call. Agents provisioned through [AWS AgentCore](/recipes/agent-integrations/agentcore-payments) can delegate wallet access to AgentCore's payment runtime while retaining user-controlled custody through Privy.

## Get started

<CardGroup>
  <Card title="Get started with agent wallets" icon="robot" href="/recipes/agent-integrations/overview">
    Provision wallets for autonomous agents
  </Card>

  <Card title="Authorization keys" icon="key" href="/controls/authorization-keys/owners/overview">
    Grant agents scoped signing access to user wallets
  </Card>

  <Card title="Set up policies" icon="shield-check" href="/controls/policies/overview">
    Define spending limits and behavioral guardrails
  </Card>

  <Card title="Agent CLI" icon="terminal" href="/recipes/agent-integrations/agent-cli">
    Spin up and fund agent wallets from the command line
  </Card>
</CardGroup>
