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

# Overview

> Pay-per-use tools for AI agents, available through one CLI.

Cralo gives AI agents access to pay-per-use tools through one CLI. Agents can discover capabilities, inspect schemas and prices, and run tools without requiring a separate provider account for every integration.

Stable JSON output makes every command easy for agents to call, parse, and compose into larger workflows.

<Info>
  The public CLI package is preparing for release. These docs describe the
  launch interface; install commands will work after `@cralo/cli` is published
  to npm.
</Info>

## Set up your AI agent

Give this prompt to your AI agent:

```text theme={null}
Set up Cralo by following https://cralo.ai/SKILL.md
```

<Columns cols={2}>
  <Card title="Get started" icon="rocket" href="/docs/quickstart">
    Equip your AI agent with Cralo and make its first tool call.
  </Card>

  <Card title="CLI reference" icon="terminal" href="/docs/cli/overview">
    See the commands, JSON contracts, input formats, and exit codes agents use.
  </Card>

  <Card title="Discover and call" icon="search" href="/docs/guides/discover-and-call">
    Let an agent search tools, inspect schemas and prices, then invoke safely.
  </Card>

  <Card title="Billing and safety" icon="shield-check" href="/docs/guides/billing-and-safety">
    Control agent spending with credits, price limits, receipts, and idempotency.
  </Card>

  <Card title="MCP server" icon="plug" href="/docs/cli/mcp">
    Run Cralo as a local MCP server for Cursor, Claude Code, and Claude Desktop.
  </Card>
</Columns>

## How it works

1. Your agent **discovers** available tools by name or capability.
2. It **describes** the selected tool to get its exact input schema and current price.
3. It **calls** the tool with JSON input. Cralo charges credits only after a successful invocation.

Agents should use `--json` so stdout contains one structured result while diagnostics remain on stderr.

```bash theme={null}
cralo tools search "extract web page" --json
cralo tools describe page.extract --json
cralo tools call page.extract --input '{"url":"https://example.com"}' --json
```

## What tools are available

The Cralo catalog is live and growing. Current capability categories include web search and research, page extraction and crawling, person and company enrichment, social media scraping (Instagram, TikTok, Reddit, X/Twitter), SERP and SEO data, and video transcription.

Browse the full catalog with live prices at [cralo.ai/tools](https://cralo.ai/tools), or discover from the terminal with `cralo tools search`.

<Tip>
  Start with [Quickstart](/docs/quickstart), then use the [CLI
  reference](/docs/cli/overview) as your command guide.
</Tip>