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

# Connect lobstr.io to Claude and run scrapers from a conversation

> Add the lobstr.io MCP server to Claude so you can find a scraper, launch a run, follow its progress, and read the results without leaving the chat.

Claude can operate your lobstr.io account directly through the lobstr.io **MCP server**. Instead of switching to the dashboard to pick a scraper, add tasks, launch, wait, and download a CSV, you describe what you want in plain language and Claude does it in your account.

**Endpoint:** `https://mcp.lobstr.io/mcp`

> [!NOTE]
>
> This is the **operational** server — it acts in your account and spends credits. There is also a separate, public documentation server at `https://docs.lobstr.io/mcp` that only answers questions about the lobstr.io API and can't touch your account.

## Before you start

* A [lobstr.io account](https://app.lobstr.io) with credits — launching a run spends credits from your balance.
* Claude on the web, Claude Desktop, or Claude Code.

You don't need your API key for this. The connection uses OAuth: you approve it on lobstr.io's own site, so **no password ever reaches Claude**.

## Watch the setup walkthrough

[https://www.loom.com/embed/e92bb83026a541558f90edf032b5fd20](https://www.loom.com/embed/e92bb83026a541558f90edf032b5fd20)

## Connect Claude to lobstr.io

> [tab] Claude.ai and Claude Desktop
>
> **1. Open your connector settings**
>
> In Claude, go to **Settings** → **Connectors**, then click **Add custom connector**.
>
> **2. Paste the lobstr.io endpoint**
>
> Give the connector a name (for example, `lobstr.io`) and paste the full URL:
>
> ```
> https://mcp.lobstr.io/mcp
> ```
>
> > [!WARNING]
> >
> > Include the trailing `/mcp`. A URL without it won't connect.
>
> **3. Approve the connection**
>
> Claude opens the lobstr.io consent page. You're already signed in, so you just review the permissions and click **Allow**. Claude returns to the conversation with the connector enabled.

> [tab] Claude Code
>
> Run this once in your terminal:
>
> ```bash theme={null}
> claude mcp add --transport http lobstr https://mcp.lobstr.io/mcp
> ```
>
> The first time you use it, Claude Code opens the lobstr.io consent page in your browser. On a headless machine, the authorize URL is printed for you to open manually.

## What you're approving

The consent page lists the permissions Claude is asking for, and you choose which ones to allow. You can revoke access at any time from your lobstr.io account.

| Permission | What it grants |
| - | - |
| `crawlers:read` | Discover scrapers and read their settings |
| `runs:read` | Check the status of your runs |
| `results:read` | Read your scraped results |
| `profile:read` | View your profile (name, email, plan) and credit balance |
| `account:read` | View your synced platform accounts (LinkedIn, Facebook, etc.) and their status |
| `runs:execute` | Run scrapers on your behalf (spends credits). Together with `account:read`, it also lets Claude attach a synced account to a Squid |

## What Claude can do

| Capability | Spends credits |
| - | - |
| Search the store for the right scraper | No |
| Read a scraper's inputs, output fields, and pricing | No |
| List and inspect your saved Squids | No |
| Create a Squid, add tasks, and change its settings without running it | No |
| Estimate a run's credits, result count, and duration | No |
| Launch a run | **Yes** |
| Check a run's status, wait for it to finish, or list a Squid's recent runs | No |
| Abort a run (results collected so far are kept) | No |
| Read a page of results, or get a download link (CSV, Excel, JSON, or JSONL) | No |
| Empty a Squid's tasks, or deactivate it (stops its schedule and any run in progress) | No |
| Check your credit balance, slots, and plan | No |
| List your synced accounts and attach one to a Squid | No |

New scrapers work automatically. Claude reads each scraper's live input schema, so nothing needs updating when lobstr.io adds one to the store.

## Scrapers that need a synced account

Some scrapers log in to a platform on your behalf, such as the LinkedIn and Sales Navigator scrapers. They need a synced account first, set up with [Account sync](https://help.lobstr.io/getting-started/account-sync.md).

* **One matching account:** if exactly one healthy synced account of the right type exists, Claude attaches it for you.
* **Several or none:** Claude asks you which account to use, or tells you which platform to sync.
* **The type must match.** A LinkedIn account can't run a Sales Navigator scraper, and the reverse.
* **Attaching only adds.** An account already on a Squid is never swapped out.

## A typical conversation

Here's a full run from start to finish — finding a scraper, launching it, and reading the results back in the chat.

[https://www.loom.com/embed/5424df66853b4f1c91a24eb8d4078aac](https://www.loom.com/embed/5424df66853b4f1c91a24eb8d4078aac)

**1. Find a scraper**

*"Find me a scraper for Google Maps business leads."*

**2. Check what it needs**

*"What inputs does it take, and what does it cost?"*

**3. Launch the run**

*"Run it for dentists in Manchester."* — Claude asks you to confirm the cost before it launches.

**4. Follow the run**

*"Is it finished?"*

**5. Get the data**

*"Show me the results."*

## Credits and confirmation

lobstr.io prices **per result row**, so the final cost of a run isn't known until it finishes. Before launching, Claude estimates the cost — and when the estimate is high or can't be calculated, it asks you to confirm instead of launching straight away. Reading scrapers, runs, and results never spends credits.

For a closer figure before anything launches, ask Claude to estimate the run once the Squid has its tasks. If email verification is on, it's billed separately after the scrape.

> [!WARNING]
>
> Runs launched from Claude are real runs in your account. They appear in your dashboard with an `mcp` badge next to the Squid name, count against your [slots](https://help.lobstr.io/core-concepts/slots.md), and draw down the same credit balance as runs you launch yourself.

## Troubleshooting

> [accordion] The connection fails or Claude can't authorize
>
> Check the URL is exactly `https://mcp.lobstr.io/mcp`, over HTTPS and with the `/mcp` path included.

> [accordion] Every run asks me to confirm
>
> That's expected. lobstr.io charges per row, so the total is unknowable up front — the estimate shows you the per-row rate so you can decide.

> [accordion] A run stays pending or errors out
>
> Open the run in your [lobstr.io dashboard](https://app.lobstr.io). Claude sees the same status your account does, so the dashboard will show the same [stop reason](https://help.lobstr.io/core-concepts/run-stop-reasons.md).

> [accordion] Claude says it has no credits or no access
>
> Confirm you clicked **Allow** on the consent page and that your balance isn't empty. You can re-approve the connector from your Claude settings.

> [accordion] Claude says I'm over my slot limit
>
> Your plan caps how many runs are in progress at the same time (its [slots](https://help.lobstr.io/core-concepts/slots.md)): Squids at rest use none. If a launch is refused because your other runs use every slot, wait for one to finish, or stop it from the [dashboard](https://app.lobstr.io) or by asking Claude; nothing is queued. Creating Squids uses no slot: your plan holds up to 100 Squids per slot.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.