Stekpad MCP for Claude Desktop.
Eight MCP tools. Reads are free. Two minutes to install.
Eight tools, free reads, typed errors.
- Claude Desktop calls Stekpad’s `scrape`, `crawl`, `map`, `extract`, `search` directly as MCP tools.
- Three free read tools (`list_datasets`, `get_dataset`, `query_dataset`) so the model can remember its own work without burning credits.
- Per-call credit reporting in every tool response.
- Typed errors with `code`, `message`, and `retry_after`.
Three steps.
- 01
Get an API key
Sign up at stekpad.com (free, no card). Go to
Settings → API keys, click Create key, name it “Claude Desktop”, copy thestkpd_live_...token. - 02
Edit the Claude Desktop config
Open Claude Desktop. Go to
Settings → Developer → Edit Config. Add the Stekpad block (or merge with your existingmcpServersobject). Save and restart Claude Desktop. - 03
Verify
Open a new chat. Click the tools menu (the small spanner icon). You should see eight tools under “stekpad”:
scrape,crawl,map,extract,search,list_datasets,get_dataset,query_dataset. Try: “Scrape https://example.com and tell me what’s on the page.” The response will includecredits_charged: 1.
{ "mcpServers": { "stekpad": { "command": "npx", "args": ["-y", "@stekpad/mcp"], "env": { "STEKPAD_API_KEY": "stkpd_live_..." } } }}Five prompts to try.
- “Search for ‘best vector databases 2026’, scrape the top 5 results, save them to a dataset called ‘Vector DB research’, and summarize each.”
- “Map the URLs on https://example.com/blog and tell me how many posts there are.”
- “Extract the company name, founders, and funding total from https://example.com/about into JSON.”
- “Show me my datasets and pick the most recent one.” (free — uses list_datasets)
- “Query the ‘Q2 outbound’ dataset for rows where the country is France.” (free — uses query_dataset)
Cookie bridge from Chrome.
If you want Claude to scrape pages behind a login (Stripe dashboard, an internal CRM, a paywalled article you have access to), install the Stekpad Chrome extension and authorize the domain. Then prompt Claude with: “Scrape https://dashboard.stripe.com/customers using my session.” Claude calls scrape with use_session: "stripe.com", the cookie bridge fetches the page in your Chrome, and the rendered content comes back. Your cookies never reach Stekpad servers. See the cookie bridge feature.
Common issues.
“Tool not found”
Restart Claude Desktop after editing the config.
“Unauthorized”
Your API key is wrong or revoked. Re-copy from Settings → API keys.
“insufficient_credits”
Top up your wallet. Reads still work.
“session_unavailable”
The cookie bridge isn’t connected. Open Chrome with the extension, log in to the target domain, retry.
`npx` not found
Install Node.js 18+ globally.
Free, forever, for personal use.
- 300 credits/month free, forever. Reads are free, always.
- Add a credit pack from 9 € when you need more.
- Cloud plans only if you want scheduling, webhooks, or team seats.
Common questions.
Does this work with Claude.ai (the web app)?
Yes via Claude Connect. Use the hosted HTTP MCP at https://mcp.stekpad.com with Authorization: Bearer stkpd_live_...
Can I limit which tools Claude has access to?
Yes. Per-tool gating is on the v1.5 roadmap. For now, use a separate API key per project and revoke when done.
How fast is a typical scrape?
1–3 seconds for a public page, 2–5 seconds via the cookie bridge.