# Build with AI

> Pick the right Alchemy tool for your AI workflow. Teach a coding agent every Alchemy API with Agent Skills, automate from the terminal with the CLI, give it live data via the MCP Server, or let it authenticate autonomously with a wallet.

> For the complete documentation index, see [llms.txt](/docs/llms.txt).

We ship four tools for building with AI on Alchemy. They're complementary, and most agent-driven workflows use two or three of them together.

## The four tools

### Agent Skills

[Alchemy Agent Skills](https://github.com/alchemyplatform/skills) are plug-and-play knowledge packages for AI coding agents. Install them once, and your agent knows every Alchemy endpoint, authentication method, pagination rule, and error pattern without manual prompting.

```bash
npx skills add alchemyplatform/skills --yes
```

* Two skills included: `alchemy-api` (API key auth) and `agentic-gateway` (wallet-based x402 auth)
* Works with [Claude Code](https://www.anthropic.com/claude-code), [Codex](https://openai.com/index/codex/), [Cursor](https://cursor.com), [VS Code Copilot](https://code.visualstudio.com/docs/copilot/overview), and any agent that supports the [skills spec](https://agentskills.io)

[Get started with Agent Skills →](docs/alchemy-agent-skills)

### Alchemy CLI

The [Alchemy CLI](docs/alchemy-cli) is a single binary that wraps every Alchemy product. Use it for live querying, admin tasks, and local automation, or give it to an AI agent as a structured shell tool.

```bash
npm i -g @alchemy/cli@latest
```

* Covers JSON-RPC nodes, Data APIs, Wallet APIs, Notify webhooks, and the Admin API across [100+ networks](docs/reference/node-supported-chains)
* `--json --no-interactive` flags make every command machine-readable
* Built-in `agent-prompt` command emits a complete usage spec for AI agents
* Supports API key, access key, and x402 wallet authentication

[Get started with the CLI →](docs/alchemy-cli)

### MCP Server

The [Alchemy MCP Server](https://github.com/alchemyplatform/alchemy-mcp-server) exposes our APIs through the [Model Context Protocol](https://modelcontextprotocol.io/), so AI tools can fetch blockchain data as tool calls inside a conversation.

```bash
claude mcp add alchemy --transport http https://mcp.alchemy.com/mcp
```

* 159 tools across token prices, NFT metadata, transaction history, simulation, tracing, account abstraction, Solana DAS, and more
* 100+ supported networks
* Hosted at `https://mcp.alchemy.com/mcp` with OAuth, no API key or local install required

[Get started with the MCP Server →](docs/alchemy-mcp-server)

### Agent authentication and payment

Your agent can authenticate with a crypto wallet instead of an API key, purchasing credits in USDC via the [x402 payment protocol](https://www.x402.org/). This is what the `agentic-gateway` skill teaches your agent automatically, and what the CLI uses when you pass `--x402`.

* No API key required, supports authentication via [SIWE (Sign-In With Ethereum)](https://eips.ethereum.org/EIPS/eip-4361) and SIWS (Sign-In With Solana)
* Supports payment via USDC on Base and Solana
* Wallet type is independent of which chain you query

[Learn about agent authentication and payment →](docs/alchemy-for-agents)

## Pick the right tool

| I want to... | Use |
|---|---|
| Give my AI coding agent knowledge of all Alchemy APIs | [Agent Skills](docs/alchemy-agent-skills) |
| Run quick onchain queries or admin tasks from the terminal | [CLI](docs/alchemy-cli) |
| Script a recurring job, cron, or CI step that calls Alchemy | [CLI](docs/alchemy-cli) |
| Give a shell-driven agent a structured tool surface | [CLI](docs/alchemy-cli) with `--json --no-interactive` |
| Let my IDE copilot query live onchain data during a conversation | [MCP Server](docs/alchemy-mcp-server) or [CLI](docs/alchemy-cli) |
| Build an autonomous agent that authenticates with a wallet and pays in USDC | [Agent Authentication and Payment](docs/alchemy-for-agents) plus [Skills](docs/alchemy-agent-skills) or [CLI](docs/alchemy-cli) |
| Give a coding agent both API knowledge and live data access | Agent Skills + MCP Server or CLI |

These tools layer on top of each other. An agent using **Skills** to understand our APIs can shell out to the **CLI** for admin tasks, call the **MCP Server** for live data, and use the **`agentic-gateway` skill** or `--x402` flag to operate without a pre-provisioned API key.

## Supported AI tools

| Tool | Skills | CLI | MCP Server | Agent Auth |
|---|---|---|---|---|
| [Claude Code](https://www.anthropic.com/claude-code) | Yes | Yes | Yes | Yes |
| [Cursor](https://cursor.com) | Yes | Yes | Yes | Yes |
| [Codex](https://openai.com/index/codex/) | Yes | Yes | Yes | Yes |
| [VS Code Copilot](https://code.visualstudio.com/docs/copilot/overview) | Yes | Yes | Yes | — |
| [Claude Desktop](https://claude.ai/download) | Yes | Yes | Yes | — |

The CLI works with any agent that can run shell commands, so the table above is a non-exhaustive list of the agents we test against.