> For the complete documentation index, see [llms.txt](https://covenant-3.gitbook.io/covenant-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://covenant-3.gitbook.io/covenant-docs/introduction/overview.md).

# Overview

> A one-page tour of what Covenant is, who it's for, and how the pieces fit. For the deeper "why," continue to [The Problem](/covenant-docs/introduction/the-problem.md) and [The Solution](/covenant-docs/introduction/the-solution.md).

## What is Covenant?

Covenant is a **safety layer for autonomous AI agents that need to spend money**. It lets an agent pay for [x402](https://x402.org) web services on its own, while guaranteeing the agent can only ever spend **inside a policy the user signed once** — the *covenant*.

Technically, a covenant is an **ERC-7710 delegation** from the user's **MetaMask Smart Account** to the agent's executor, carrying on-chain **caveats** (a USDC budget cap and an expiry) plus an off-chain policy (allowed services, purpose, per-request cap, and more). The agent pays by **redeeming** that delegation; a **policy firewall** vets every payment first.

## Who is it for?

| Audience                       | Why they care                                                                         |
| ------------------------------ | ------------------------------------------------------------------------------------- |
| **Users delegating to agents** | Sign once, then let an agent work unattended with a hard guarantee on the worst case. |
| **Agent / framework builders** | A drop-in pattern for safe, autonomous, non-custodial payments.                       |
| **x402 service providers**     | A paying client whose payments are verifiable on-chain.                               |
| **Hackathon judges**           | A substantive composition of x402 + ERC-7710, not two checkboxes.                     |

## How the pieces fit

* **MetaMask Smart Account** — holds the user's funds; signs the covenant. Nothing is custodied by us.
* **ERC-7710 delegation + caveats** — the cryptographic budget and expiry the agent cannot exceed.
* **Policy engine** — the off-chain firewall enforcing intent-level rules before any payment.
* **x402 service** — the paid API; issues a `402` envelope and verifies payment on-chain.
* **Venice AI** — the agent's planner and report writer.
* **Base Sepolia + USDC** — the chain and the settlement asset.

```mermaid
flowchart LR
  SA["MetaMask Smart Account"] -->|signs| COV["Covenant (ERC-7710)"]
  COV --> CAV["On-chain caveats: budget + expiry"]
  AG["AI Agent"] --> PE["Policy engine"]
  X["x402 service"] --> PE
  PE -->|approved| RD["redeemDelegations"]
  CAV -. guards .-> RD
  RD --> USDC[("USDC on Base Sepolia")]
  VEN["Venice AI"] -. plans & reports .- AG
```

## The guarantee, in one sentence

> **Worst case, the agent spends the budget you set, before the time you set, and nothing more** — even if it is fully compromised.

## Where to go next

* New to the space? → [Background](/covenant-docs/introduction/background.md)
* Want the motivation? → [The Problem](/covenant-docs/introduction/the-problem.md) · [The Solution](/covenant-docs/introduction/the-solution.md)
* Want the mechanics? → [How It Works](/covenant-docs/core-concepts/how-it-works.md)
* Want to run it? → [Quickstart](/covenant-docs/getting-started/quickstart.md)


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://covenant-3.gitbook.io/covenant-docs/introduction/overview.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.