A Connector So Claude and ChatGPT Can Use Your Product
An MCP server built over your existing API, for SaaS and dev-tool teams. Four to six read-only tools, one sign-in flow, deployment to your infrastructure, and tests in the clients your users run.
$1,190 fixed, 15 business days. If you've paid for the $390 connector review, it counts toward this build and you pay the remaining $800.
Who this is for
Your customers want to ask Claude or ChatGPT questions about your product, or your own team wants to work with your API through AI clients. You need a connector your team owns, shaped around real user questions and built to published review criteria.
It's a fit when you have a documented API, a populated test account and a deployment target. It's not a fit for financial-asset transfers or AI-generated images, video or audio, because Anthropic's connector criteria don't accept them.
What you get
- 1.Four to six read-only tools. Each tool calls a fixed API operation and has a clear name, a title and truthful annotations. Read and write operations stay separate, in line with Anthropic's published tool-design criteria.
- 2.One sign-in flow. Your existing OAuth flow, or an authorization layer around an API-key product. Tokens go in authorization headers, not in URLs.
- 3.A remote HTTPS server plus local mode. Delivered as a Docker image and deployed once to your infrastructure.
- 4.Tests and client checks. Unit and integration tests for your CI, an MCP Inspector log, and every tool exercised in Claude, Claude Code, Codex and ChatGPT developer mode, where your account and hosting support them.
- 5.A setup page for your users. Installation, authentication, example questions and troubleshooting.
- 6.A submission pack. The materials your team needs to file with Anthropic and OpenAI under its own accounts. You keep control of the submission and the listing.
- 7.A recorded handover. A 45-minute walkthrough of the code, tests, deployment and maintenance.
Built in TypeScript or Ruby, so your developers can read and maintain it.
What the fixed price covers
| Item | Included |
|---|---|
| Four to six read-only tools over one existing API | Yes |
| One authentication approach | Yes |
| One deployment target | Yes |
| Unit and integration tests | Yes |
| MCP Inspector and supported client checks | Yes |
| User documentation and submission pack | Yes |
| Recorded handover and 30-day fix window | Yes |
| Fixed build price | $1,190 |
| Independent connector review | $390, credited toward the build |
| Build balance after a paid review | $800 |
Extra tools, write actions, another authentication approach or another deployment target are scoped and quoted in writing before work starts. Fixes for issues found in a separate review aren't included unless we agree a build or fix scope.
How it works
- Step 1
A 20-minute call to choose the user workflows and the sign-in approach.
- Step 2
You get a proposal with a start date, then send API documentation, a populated test account and credentials.
- Step 3
The 15 business days start on the agreed start date or the day your access list is complete, whichever comes later.
- Step 4
Day 5: you try the tools against a staging URL and send comments.
- Step 5
Day 10: deployment, green tests, documentation, submission pack and handover.
- Step 6
For 30 days, I fix anything that misses the agreed acceptance list.
Review criteria and directory listings
The build follows the relevant parts of Anthropic's connector pre-submission checklist: separate read and write tools, accurate names and descriptions, applicable annotations, functional error handling, API ownership and complete test materials.
Following the checklist doesn't mean Anthropic approves or certifies the connector. Anthropic scans submissions automatically, lists Community connectors by default and may escalate some listings to Verified review. Your team submits under its own account, and neither a listing nor a Verified label is guaranteed.
Price and payment
$1,190 fixed. 50% at kickoff and 50% on acceptance, by invoice, paid by card or bank transfer.
Already bought the $390 connector review? It counts toward the $1,190 build, leaving $800 to pay.
Questions
Will our connector be listed in the Claude directory?
Anthropic decides, and nobody can promise a listing or a Verified label. Your team submits from its own account. I build to the published criteria and prepare the materials the submission asks for.
Does it work in ChatGPT?
Yes, as a custom connector in developer mode, where your plan supports it. For a public listing, the pack also covers OpenAI's plugin submission process. OpenAI makes that decision too.
Can it write data as well as read it?
Yes, scoped and quoted separately. Write tools come with confirmation behavior and destructive annotations. Most teams start read-only.
Who hosts the server?
You do. The build includes one deployment to your target and a runbook for your team.
What do you need from us?
API documentation or an OpenAPI file, a populated test account, authentication credentials or an API key, three example user questions, your logo, a support email, and your privacy policy and terms URLs.
What if our API has no OAuth and no API keys?
The connector needs an authentication service first. I quote that separately after the call.
Do you use AI tools on our code?
Yes. I build with Claude Code and Codex, and a second model reviews commits before they merge. We agree access, data handling and account terms before work starts. I don't use production data for development or testing.
Can we start with a review instead?
Yes. The independent connector review is $390 and follows Anthropic's published criteria. If you go ahead with the build, the full $390 counts toward the $1,190.
Still have a question? Ask me directly.
Three ways to start
- 1. Two Engines Kit: the Claude Code and Codex workflow used to build and review this work, for teams that want to run it in their own repositories.
- 2. MCP connector review: an independent review for teams that already have a connector. $390, credited toward the build.
- 3. MCP server build: a connector built, tested and documented for your existing API. $1,190 fixed.
Sources
- Anthropic: Connector pre-submission checklist. Fetched 5 Nov 2026. Source for tool separation, API documentation in custom query tools, annotations, name length, descriptions, prompt-injection checks, functional quality, API ownership, unsupported use cases, submission materials, testing and review labels.
- OpenAI: Plugin guidelines. Source for OpenAI controlling plugin acceptance and requiring verified submitters.
- OpenAI: Plugin submission. Source for the plugin submission process and materials.
- RayaStats: How to make Codex review Claude Code commits. Source for the described Claude Code and Codex build-and-review workflow.