> For the complete documentation index, see [llms.txt](https://docs.pepeunchained.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.pepeunchained.com/get-started/ai-assisted.md).

# Build with an AI Coding Agent in 3 Minutes

The fastest path for agent users: install the MCP server, the plugin and the skills, then scaffold and verify an app without leaving the agent. The agent does the code and config; you provide the Privy credentials. This page is the short path; detail is in the agent-tooling and MCP sections.

![Dev kit map: the SDK at the base; the MCP server with 33 tools, 12 skills, the Claude Code plugin, the create-pepu-app CLI and 6 templates all build on it](https://3849619095-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FKvVTxzL1E48r5cDRL2er%2Fuploads%2Fgit-blob-124a3bf8f925ffa6013a7f99ea2b11897bcc4b11%2Fdevkit-map.png?alt=media)

*Everything builds on the SDK.*

The MCP server (`@pepu-chain/mcp`: 33 tools, 6 prompts), 12 skills, 6 commands and the builder subagent (plugin 0.4.4) cover the whole dev kit. Gas sponsorship uses PEPU's hosted bundler and paymaster and needs a funded project of yours.

## Use it with an AI agent: the fast path (Claude Code)

Three steps, about three minutes of your time.

1. Install the plugin (once):

   ```
   /plugin marketplace add pepu-org/pepu-devkit
   /plugin install pepu-devkit@pepu
   ```
2. In an empty folder, run the command with your idea:

   ```
   /pepu-new-app a tap game where players pay no gas and approve once
   ```
3. Answer the agent when it asks for the few things only you have (below), then run what it tells you: `npm run dev`.

What `/pepu-new-app` does, in order:

| Step | What the agent does                                                                                                          | Tool                      |
| ---- | ---------------------------------------------------------------------------------------------------------------------------- | ------------------------- |
| 1    | Lists the templates and picks one by what you described, saying why in one line. For the idea above: `sponsored-session-app` | `pepu_list_templates`     |
| 2    | Scaffolds the app into a new folder (the same as `npx create-pepu-app`), then installs dependencies                          | `pepu_scaffold_app`       |
| 3    | Reads `.env.local`, fills what it can, and asks you for the rest by name without echoing secrets                             | `pepu_check_env`          |
| 4    | Checks your Privy dashboard for the dev origin and tells you which rows to change                                            | `pepu_check_privy_config` |
| 5    | For a sponsored template, checks the hosted bundler and paymaster live and says what is missing (your project's key)         | `pepu_sponsorship_status` |
| 6    | Runs typecheck and build and ends with the three commands to run                                                             |                           |

What only you do:

1. Create the Privy app at [dashboard.privy.io](https://dashboard.privy.io) and give the agent the App ID and App Secret (or paste them into `.env.local` yourself).
2. Click the dashboard settings the agent lists: Allowed origins `http://localhost:3000`, Google on, embedded wallets on, and Smart Wallets only if it picked `sponsored-demo`. The exact clicks are in [Dashboard Settings](/privy/dashboard-settings.md).
3. For a sponsored template, get a paymaster project with its key. Ask: "Walk me through claiming and funding a paymaster project for my Privy app." The agent encodes the transactions with `pepu_prepare_project_tx`; you sign them with the owner wallet and run the SIWE and attestation script, which needs your wallet and your Privy secret. See [Register, Deposit and Claim a v2 Project](/infrastructure/v2-register-deposit-claim.md).
4. Run `npm run doctor`, then `npm run dev`, and click through the flow. When it looks right, the agent can re-run `/pepu-doctor` for a readiness report.

Other clients (Cursor, VS Code, Windsurf, Claude Desktop, Cline) get the MCP server only, no slash commands: use the first prompt below. The rest of this page is the manual route.

## Prerequisites

* Node.js 18.17 or newer and an MCP-capable agent client.
* A Privy account (you create the app and copy two values).

## 1. Install the MCP server

Every client takes the same stdio server:

```bash
npx -y @pepu-chain/mcp
```

Claude Code, Cursor, VS Code, Windsurf, Claude Desktop and Cline each have their own config file shape. The shared entry is:

```json
{ "mcpServers": { "pepu": { "command": "npx", "args": ["-y", "@pepu-chain/mcp"] } } }
```

Exact per-client steps: [Install per Client](/mcp/install.md). Scaffolded templates already ship a `.mcp.json`. No API key is needed. Optional env: `PEPU_BUNDLER_URL`, `PEPU_PAYMASTER_URL`, `PEPU_RPC_URL`.

## 2. Claude Code plugin

```
/plugin marketplace add pepu-org/pepu-devkit
/plugin install pepu-devkit@pepu
```

It adds the MCP server, 12 skills, a builder subagent and six commands:

| Command                       | Does                                                    |
| ----------------------------- | ------------------------------------------------------- |
| `/pepu-new-app`               | Scaffolds an app, choosing the template by pillar       |
| `/pepu-add-login`             | Google login and an embedded wallet                     |
| `/pepu-add-session`           | Privy session keys                                      |
| `/pepu-add-sponsorship`       | Sponsored transactions, after a live availability check |
| `/pepu-add-sponsored-session` | The full combination                                    |
| `/pepu-doctor`                | Env, Privy dashboard and sponsorship readiness          |

See [Claude Code Plugin](/agent-tooling/plugin.md) and [Skills Catalog](/agent-tooling/skills.md).

## What the agent does unattended, and what needs you

| Agent can                                                            | Needs a human                                                                                                            |
| -------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------ |
| Scaffold the app (`pepu_scaffold_app`)                               | Create the Privy app at dashboard.privy.io                                                                               |
| Write SDK code, routes and hooks                                     | Copy the App ID and App Secret into `.env.local` (or paste them to the agent; the tools never echo secrets)              |
| Generate the authorization key into `.env.local` without printing it | Register the public key and create the key quorum in the Privy dashboard                                                 |
| Check env keys, Privy config, sponsorship status                     | Set Allowed origins, enable Google, enable embedded wallets                                                              |
| Run `npm run doctor`, typecheck, build                               | Provide the paymaster project key for sponsored templates (`PEPU_PAYMASTER_URL`; the bundler defaults to the hosted one) |
| Explain a failed transaction or error                                | Click through the browser flow and approve Privy prompts                                                                 |

## A first prompt

```
Create a PEPU app that has Google login and a wallet. Use the pepu MCP tools and the
pepu-build-dapp skill. Scaffold it in ./my-app with the consumer-app template, tell me
exactly which Privy dashboard settings I must change, then run npm run doctor and
pepu_check_privy_config with origin http://localhost:3000 and report what is left.
```

For no-popup, no-gas, ask for the `sponsored-session-app` template and say you will supply a paymaster project token (the bundler URL defaults to the hosted one).

## What to expect

The agent lists the template choice with one line of reasoning, scaffolds, reports missing env keys by name, and finishes with a doctor report. Fields Privy does not expose come back `unknown`, not failed.

## Common failures

| Symptom                            | Cause                                          | Fix                                   |
| ---------------------------------- | ---------------------------------------------- | ------------------------------------- |
| Tools not listed                   | MCP not registered or the client not restarted | [Install per Client](/mcp/install.md) |
| Doctor: `PRIVY_APP_SECRET` missing | You have not provided it                       | Add it to `.env.local`, restart dev   |
| Agent claims sponsorship is live   | Skipped the live check                         | Run `pepu_sponsorship_status`         |

## Related

* [MCP Server Overview](/mcp/index.md)
* [MCP and Skills overview](/agent-tooling/mcp-and-skills.md)
* [Quickstart: Zero to a Sponsored Transaction in 10 Minutes](/get-started/quickstart-app.md)
* [Smart Account Paths](/concepts/smart-account-paths.md)
* [AGENTS.md and CLAUDE.md in Templates](/agent-tooling/agents-md.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 by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.pepeunchained.com/get-started/ai-assisted.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

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.