MCP servers, hosted and sold

An MCP server is a tool an AI client such as Claude, ChatGPT or Cursor can call: it publishes a schema, checks who is calling and logs every call. GTDataWorks is bringing its own to the MCPize marketplace and hosts yours, private or listed for sale.

One tool call, start to finish

AI clientClaude, ChatGPT, Cursor or any MCP-capable client
MCP serverPublished schema, API key checked, every call logged
Your data or APIDatabase, documents or an internal service
ReceiptWho called, which tool, when, and a hash of the result

Every call leaves a receipt, and the receipts are what you audit later.

GTDataWorks on MCPize

Two servers are being packaged for the marketplace now. Neither is live yet; each listing opens with its schema and its pricing.

terminus-witness

Listing in preparation

A provenance witness. It hashes and timestamps what an agent produced and appends the entry to a receipt chain, so a third party can verify later that the output existed at that time and has not changed since.

Schema and pricing will be published with the listing.

terminus-admission

Listing in preparation

Admission control for agent actions. Before an agent's intended action runs, the server checks it against your policy, allows or refuses it, and records the decision.

Schema and pricing will be published with the listing.

Apify Actors

Two Actors for the Apify platform, packaged alongside the servers and also in preparation.

ActorWhat it doesStatus

gtd-refinery

Dataset refinery for the GTDataWorks dataset factoryIn preparation

receipt-fixer

Turns a folder of receipts into a verified CSVIn preparation

Host your MCP server

The managed service: your server packaged, hosted, metered and kept current, so your team or your customers can call it without running infrastructure.

What is included

  • Hosted endpoint on Modal or Google Cloud Run, packaged in Docker
  • An API key for each caller, metered per key
  • Updates on a set cadence
  • Logging and a usage receipt for every call
  • Monthly usage report

Private or listed

The server can stay private to your team or your customers, or be listed for sale on MCPize. The deliverables are the same either way: a hosted endpoint, key management, a monthly usage report and an update cadence.

Timeline
1–2 weeks to launch
Runs on
Modal or Google Cloud Run
Access
API keys, one per caller

Setup

$1,000–$2,000

Packaging, hosting, key management and launch.

High volume

per call

A per-call model is available for high-volume servers. Ask for it in the brief.

Cost estimate. The market review this site draws on puts the compute for a small LLM-connected server at roughly $50–$200 a month, depending on traffic. That is an estimate from the report, not a measured bill; the quote for your server follows a scoping call.

How a server gets to market

Five steps from a working server to a metered endpoint with a monthly report.

  1. Package the server

    MCP schema published, authentication in place, tests passing. The package is what gets hosted and what buyers see.

  2. Add metering and keys

    Each caller gets an API key. Every call is counted against that key and logged with a receipt.

  3. Choose the route

    A private endpoint for your own team or customers, or a public listing on MCPize.

  4. Launch

    The endpoint goes live on Modal or Google Cloud Run. From brief to launch is one to two weeks.

  5. Report and update

    A monthly usage report, updates on a set cadence, and logs you can audit.

Teams without a server start one step earlier: the custom LLM connectors and MCP tools service builds the connector or MCP server with tests, an access policy and an install guide, from $1,000.

Who this is for

Teams that want a server as a service

You have a database, a document set or an internal API that a model should be able to use, and nobody to run the server. GTDataWorks packages and hosts it; your team gets an endpoint, keys for each caller and a usage report every month.

Tool builders who want to sell one

You wrote an MCP server and want it on MCPize without building hosting and metering yourself. GTDataWorks runs the endpoint and meters every key, so each call is logged with a receipt and the monthly usage report shows usage per key.

Put a server on the market

Describe what the server exposes, who should be able to call it, and whether it stays private or goes on MCPize. GTDataWorks replies with a scope and the next step.