Rupesh Kumar: I connect your SaaS product to ChatGPT and Claude
Your customers increasingly work inside AI assistants. I build the MCP server that lets them use your product from there, safely, without your team pausing the roadmap.
Customers ask "can I use you from Claude?" and most small SaaS teams have no answer. Building it well means getting OAuth, per-user permissions, rate limits and logging right, which is a side project nobody on a 10-person team has time for.
What you get
Read-only pilot (1–2 weeks): an MCP server over your existing public API exposing a handful of safe, read-only tools, working in Claude and ChatGPT, with a short demo video you can show customers.
Full build (4–6 weeks): write actions, OAuth sign-in with your existing accounts, per-user scopes, an audit log of every AI action, rate limiting, docs, and a listing in the public MCP registries.
Optional upkeep: monthly maintenance as your API and the AI clients change.
How it works
A free 2-minute demo of your product inside Claude, built from your public docs.
A 30-minute call to pick the first tools and agree scope.
The pilot, delivered in your repo or as a hosted server you own.
If it's useful, the full build with auth and audit logging.
Pricing
Pilots from $1,500. Full builds are quoted after the pilot based on the number of tools and auth requirements. Fixed price, 50% upfront.
FAQ
Do you need access to our codebase?
Not for the pilot. It runs on your public API with a test account. The full build can live in your repo.
Who owns the code?
You do, fully, once paid.
Is it safe to let an AI act in our product?
The pilot is read-only. Write actions in the full build use each user's own permissions, need explicit scopes, and are logged.
Which AI clients does it work with?
Anything that supports the Model Context Protocol, including Claude and ChatGPT.