AgentToolWorks MCP servers
Remote Model Context Protocol servers that AI agents call mid-task, billed per call from one prepaid credit balance. Operated by ECOM FR LLC. Website: agenttoolworks.com.
This repository holds the public registry manifests (server.json) for each server. The servers are hosted; there is nothing to install or run locally.
Connect
- Get a key with 100 free credits, no card: agenttoolworks.com/signup. One key works on every server.
- Add a server to your client. Claude Code:
claude mcp add --transport http verifydesk \
https://verifydesk.agenttoolworks.com/mcp \
--header "Authorization: Bearer atw_live_your_key"
Cursor, Windsurf, VS Code and Claude Desktop each use a slightly different config shape; the exact snippet for each is in the docs.
Tools and prices
1 credit = $0.01 at the Starter rate. A call that fails on our side is billed zero. get_usage is free on every server.
Prepaid packs: $10 for 1,000 credits, $45 for 5,000, $80 for 10,000. No subscription, credits never expire. Full pricing.
Also on the Apify Store
Two of our data products also run as Apify Actors (the Store Catalog readers are the same code as its MCP server), paid per result through Apify, no AgentToolWorks account needed. AI agents can call them through Apify's hosted MCP server.
MCP clients can load either one with https://mcp.apify.com?tools=agenttoolworks/shopify-woocommerce-scraper (an Apify token is required).
Reliability
Every server is measured daily against production by an agent scenario suite, and the results are published as measured, including degraded runs: agenttoolworks.com/status.
Data sources
Official public sources only, listed per tool on each server page. Sanctions screening is a due diligence signal, not a legal determination. InvoiceForge checks the rules it lists and does not transmit invoices; it is not an accredited e-invoicing platform.