# Claude Code (/docs/claude-code)

Give Claude Code current public social media, commerce, and review data with a one-line install. Documented endpoint caches apply.

Add one line to your terminal and Claude Code can pull current public data from 44 bundled platforms, subject to each endpoint's documented cache window, with no HTTP plumbing to write.

  Two transports, same seven tools. The hosted server needs nothing installed;
  the local one runs the `socialcrawl-mcp` package through `npx`.

## Prerequisites

- Claude Code installed and signed in.
- A SocialCrawl API key from [socialcrawl.dev](https://www.socialcrawl.dev), under Dashboard → API Keys. Every account starts with 100 free credits, no card required.
- For the local transport only: Node.js with `npx` available.

## Server URL

```text title="MCP URL"
https://mcp.socialcrawl.dev/mcp
```

## How do I give Claude Code access to social media data?

### Register the server

Hosted, which is the shorter path:

```bash title="Terminal"
claude mcp add --transport http socialcrawl https://mcp.socialcrawl.dev/mcp \
  --header "Authorization: Bearer sc_your_key_here"
```

Or local, if you would rather run the package on your own machine:

```bash title="Terminal"
claude mcp add --scope user socialcrawl -e SOCIALCRAWL_API_KEY=sc_your_key_here -- npx -y socialcrawl-mcp
```

Flags go before the `--` separator; everything after it is the command Claude Code runs. `--scope user` makes the server available in every project.

### Restart Claude Code

The server list is read at startup, so quit and reopen before you expect the tools to appear.

### Ask for something

```text title="Prompt"
Get the TikTok profile for @charlidamelio
```

```text title="Prompt"
Search YouTube for machine learning tutorials
```

Claude Code picks the platform and endpoint, calls the API, and reads back structured JSON. Start with a discovery question ("which SocialCrawl platforms are available?") to confirm the connection for free.

Using Claude Desktop, Cursor, VS Code, or Windsurf instead? The per-client configuration blocks are on [Skills & MCP](/docs/skills-and-mcp.md).

## What tools does the agent get?

Seven, in four groups.

| Group     | Tools                                                                              | Cost                                       |
| --------- | ---------------------------------------------------------------------------------- | ------------------------------------------ |
| Discovery | `socialcrawl_list_platforms`, `socialcrawl_list_endpoints`, `socialcrawl_get_docs` | 0 credits, no API key needed               |
| Data      | `socialcrawl_request`                                                              | The endpoint's own price, 0 on a cache hit |
| Account   | `socialcrawl_check_balance`                                                        | 0 credits                                  |
| Stateful  | `socialcrawl_monitors`, `socialcrawl_web`                                          | The endpoint's own price                   |

The three discovery tools read bundled local data, so the agent can list platforms, inspect an endpoint's parameters, and read the pricing table before it spends anything. Before every billed call, the server validates locally that the platform, endpoint, and required parameters exist, and tells the agent how to fix a bad call instead of burning a credit.

For the full tool table, see [Skills & MCP](/docs/skills-and-mcp.md) or the [MCP server page](/mcp-server).

## What does it cost?

Credits are billed exactly as the REST API bills them: most endpoints are 1 credit, heavier ones 5 or 10, and composite or bundle endpoints carry their own price (`search/everywhere` is a flat 20). Cache hits cost 0 credits, and so do the discovery and documentation tools. See [Endpoint pricing](/docs/endpoint-pricing.md) for the exact figure per endpoint, [Credits](/docs/credits.md) for how the ledger works, or [Pricing](/pricing) for credit packs.

## Troubleshooting

Claude Code reads its server list at startup. Quit and reopen it after running `claude mcp add`. If you added the server inside one project without `--scope user`, it is only available in that project.

The key is not reaching the server. On the hosted transport, check that the `--header` argument was quoted as a single value: `--header "Authorization: Bearer sc_your_key_here"`. On the local transport, the key must be passed with `-e SOCIALCRAWL_API_KEY=...` before the `--` separator, because the server reads it from its own environment and never falls back to a key of its own.

The published package bundles 44 platforms and 357 endpoints, while the REST API serves 65 platforms and 572 endpoints, so the newest additions land in the package a release behind. Call the REST endpoint directly in the meantime.

```bash title="cURL"
curl https://mcp.socialcrawl.dev/healthz
```

A response means the server is up and the problem is local configuration.

## Next steps

- [Skills & MCP](/docs/skills-and-mcp.md): The full tool list, every client config, and the Agent Skill.
- [AI agent integration](/docs/ai-agents.md): Machine-readable llms.txt, OpenAPI, and per-platform reference files.
- [Vercel AI SDK](/docs/vercel-ai-sdk.md): Building your own agent instead of using an MCP client.
- [Quickstart](/docs/quickstart.md): Your first authenticated request, start to finish.