# For agents

Everything an agent needs to use UpRock: the MCP server, the skill, the CLI and the files to read. This page is also at https://uprock.ai/agents.md.

## One-paste setup

Paste this into Claude Code, Codex, Cursor, OpenClaw, Hermes or any agent with a terminal:

```text
Set up UpRock for me: read https://uprock.ai/agents.md, connect the UpRock MCP server, and run one test crawl. Tell me what it costs before you spend anything.
```

## The MCP server

Streamable HTTP. Get a key at https://uprock.ai/app/api-keys and send it as a header:

```text
URL:    https://mcp.uprock.ai/mcp
Header: Authorization: Bearer <API_KEY>
```

A client that can't send headers can put the key in the URL instead: `https://mcp.uprock.ai/<API_KEY>/mcp`. The header keeps the key out of URLs, logs and shared configs, so use it when you can.

### Connect

Claude Code:

```bash
claude mcp add --transport http uprock https://mcp.uprock.ai/mcp --header "Authorization: Bearer $UPROCK_API_KEY"
```

Cursor, Claude Desktop and other clients (`mcp.json`):

```json
{
  "mcpServers": {
    "uprock": {
      "url": "https://mcp.uprock.ai/mcp",
      "headers": { "Authorization": "Bearer <API_KEY>" }
    }
  }
}
```

No `headers` field in your client? Use `{ "url": "https://mcp.uprock.ai/<API_KEY>/mcp" }`.

Hermes Agent, OpenClaw and others: add `https://mcp.uprock.ai/mcp` as a remote (HTTP) MCP server with the header, or the key-in-URL form if the agent can't set headers.

### Tools

| Tool | What it does | Key inputs |
|---|---|---|
| `crawl_fetch` | Fetch a URL on a real device and return markdown + HTML | `url`, `country`, `device_type` (desktop, mobile) |
| `web_research` | Search several engines from one or more countries, deduplicated | `query`, `suggested_countries`, `max_results` |
| `verify` | Verify a site: load a URL on real devices in many regions; status, timings, screenshots (`sweep` still works as an alias) | `url`, `regions`, `tries` |
| `resource_fetch` | Read a `crawl://` or `sweep://` result in full | `uri` |
| `video_search` | Find moments in public video by meaning | `query`, `published_after` |
| `video_moment` | Describe what happens in a time range of a video | `url`, `start_s`, `end_s` |

Countries: 60+ ISO codes, or regions NA, EU, APAC, LATAM, MEA. Prefer regions unless you need one exact country.

## The skill

A ready-made skill that teaches an agent when and how to use UpRock.

- Download: https://uprock.ai/skill/uprock-skill.zip
- Read it: https://uprock.ai/skill/SKILL.md

Claude Code: unzip into `~/.claude/skills/` (so you have `~/.claude/skills/uprock/SKILL.md`). Claude apps: upload the zip as a custom skill.

## The CLI

```bash
npm i -g uprock
uprock auth login --email you@example.com --setup-ai   # or: export UPROCK_API_KEY=<key>
uprock ai crawl https://example.com --country DE --content
uprock ai research "best street food" --countries IN,BR
uprock ai verify https://example.com -r NA,EU,APAC
uprock ai buy 20          # Stripe link to add $20 of credits; waits for payment
```

Every command prints JSON. Flags: https://github.com/uprockcom/cli

## Files to read

| File | What it is |
|---|---|
| https://uprock.ai/llms.txt | Index of every page, one line each |
| https://uprock.ai/llms-full.txt | Everything in one file: API, pricing, FAQ |
| https://uprock.ai/agents.md | This page |
| https://uprock.ai/skill/SKILL.md | The skill |

Any page on uprock.ai returns markdown when requested with `Accept: text/markdown`.

## Pricing

1 page = 1 credit. Crawl $0.50 per 1,000 pages by card or USDC, $0.40 with UPT. Web research $1 per 1,000 queries. Video search $1 per 1,000. UPT is always 20% off. No free tier, no plans. The only free credits come from running a node: `npm i -g uprock`, `uprock auth login`, `uprock daemon install`.