A SaaS founder put his whole product inside Claude and ChatGPT with a 44-tool MCP — every small SaaS without an agent front door is a client
Howseen's founder shipped a production MCP server with 44 tools so customers can read and act on their data from Claude, ChatGPT, Cursor or Codex. He published the code that keeps writes safe. Most small SaaS still has no agent front door.
Published Nov 7, 2026
Free check · public files only
Is your store or SaaS agent-ready?
Paste a public homepage. We read a few public files — robots.txt, llms.txt, markup, and whether a pricing page or agent endpoint is in the open — and show a short score right away. The full report unlocks when you leave an email. That signup is the same newsletter as everywhere else on this site: we email a confirmation link, and you confirm on our site.
0 passed · 0 warnings · 0 failed
See the full report
Enter your email to unlock the check-by-check report and a matching Build This Business page. We only send the standard confirmation email.
Free. Double opt-in: we email you a link; tap it, then tap Confirm on our site. Unsubscribe any time. Privacy.
Full report
This is a snapshot of public files, not a certification, a rank, or a prediction of sales. We do not log in, submit forms, call any tool on an MCP server, or store the page content. Results can be cached for 24 hours per domain.
The Wow
What someone built
A stateless remote MCP at app.howseen.ai/api/mcp (v1.3.0): 26 read tools, 18 write tools and three ready-made prompts, listed in the official MCP registry. We listed the 44 tools with one keyless request on Nov 7, 2026.
- License
- Write-up and code excerpts are the author's; quote and link, and write your own implementation from the described pattern. The MCP SDK itself is open source
- What you need
- curl to list the tools (no key needed)
- A Howseen account and API key to call tools
- An MCP client such as ChatGPT, Claude, Codex or Cursor
- Install
- curl -s -X POST https://app.howseen.ai/api/mcp -H 'Content-Type: application/json' -H 'Accept: application/json, text/event-stream' -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'
- A business you could build on it
- MCP studio for small B2B SaaS
Why it's remarkable
Discovery is open so directories can list it, but every call needs a key, keys are read-only by default, and every write returns a preview that runs only with confirm: true — enforced in code, not just the prompt.
The Opportunity
The business it reveals
VEED's CEO says his agent found a tool, used it daily and asked him to pay (reported). Every SaaS with an API but no MCP server is a client for a studio that builds the same front door.
Who would pay
Small B2B SaaS founders with a public API; later, agencies that want client reports built from an assistant.
The Flip
3 ways you could turn this into a business
MCP studio for small B2B SaaS
- Who pays
- SaaS founders with a public API and no engineering time for MCP
- How it would earn
- A fixed build fee for a registry-listed server with safe writes, then a monthly keep-alive (our hypothesis)
- Why it might work
- The Howseen write-up is a working blueprint, and every listed server is a public demo you can show a prospect.
- What already exists
- Building in-house; generic API-to-MCP generators that skip the preview gate and registry listing.
Agent front-door audit for SaaS
- Who pays
- SaaS marketing and product leads who want AI assistants to recommend and use their product
- How it would earn
- A one-time report: can an agent discover the product, read its docs, sign up and call it, with a fix list (our hypothesis)
- Why it might work
- VEED's CEO says his agent found a tool on its own and asked him to pay (reported); founders want to be that tool.
- What already exists
- SEO and AI-search agencies that measure mentions but never test agent signup or tool calls.
Client-report MCP for agencies
- Who pays
- Marketing and SEO agencies that send monthly client reports
- How it would earn
- Setup plus monthly for an MCP over the agency's data so account managers build reports from Claude or ChatGPT (our hypothesis)
- Why it might work
- Howseen already ships a client_report prompt; agencies repeat that job every month.
- What already exists
- Dashboard tools with PDF exports; manual spreadsheet reports.
Why we'd try "MCP studio for small B2B SaaS" first: The buyer already has an API and a reason to care, the pattern is published, and the finished server is its own sales demo in the registry.
Build It
How you could build it
A stateless remote MCP server on a SaaS company's existing API: open discovery, keyed calls, read-only keys by default, preview-then-confirm writes, a registry entry and setup pages for each assistant.
First step: Run the tools/list curl above, read the writeGate section of the write-up, then pick one SaaS you use that has a public API and sketch the five read tools its customers would ask an assistant for.
Agents and tools
- MCP TypeScript SDK
- The client's REST API
- ChatGPT, Claude, Codex, Cursor or Gemini CLI for testing
- The official MCP registry
Setup
- List the client's top ten customer questions and map each to one read API call.
- Build the server stateless, with a discovery allowlist and the batch rule from the write-up.
- Add write tools only behind a write scope and a preview that needs confirm: true.
- Test in at least two assistants and publish a registry entry plus setup docs.
- Hand over with a keep-alive plan for spec and client changes.
What the agent does
- Drafts tool schemas and handler code from the API docs
- Writes setup docs for each assistant
- Runs test conversations against the server
What you do
- Sells to SaaS founders and scopes which data is safe to expose
- Reviews every write tool and preview message
- Handles keys, security questions and maintenance
Your first customer: A SaaS founder in your network with a public API, offered a free five-tool read-only demo in exchange for a case study.
Test demand first: Run the prompt on ten SaaS products with public APIs this week and build a five-tool read-only demo for the best one. Success: its founder asks what a full server would cost.
Your starter instructions
Paste this into the AI agent you use (ChatGPT, Claude, Gemini, Grok Bot or another) as your first message. Copying it starts work in your own agent; it doesn't create or deploy a business by itself.
Running it
What stays human: Selling to SaaS founders, mapping which API calls are safe to expose, reviewing every write tool's preview text, testing in each assistant, and handling key and security questions.
- Startup cost (estimate)
- Under $25 to build a demo (estimate) (Estimate: the MCP SDK is free, a serverless route runs on a free tier, and you test with an AI plan you may already pay for)
- Running cost
- A few dollars a month of hosting per client server plus maintenance hours (estimate) (Estimate: stateless JSON responses are light; hours unmeasured. Spec and client changes are the real keep-alive cost)
- How much the agent runs
- Agents run one core part; people run the rest.
Pricing your own version? Use the free startup-cost worksheet.
The evidence
What the source shows: Howseen's DEV write-up (Nov 6, 2026), the live tools/list response (44 tools, 26 read-only) and a 401 on a keyless tool call, plus the registry entry ai.howseen/howseen-mcp — all checked Nov 7, 2026. No revenue is published.
What's ours: The MCP studio, agent front-door audit and agency report flips are our applications of the pattern.
Worth planning around
- Howseen does not say how many users connect the MCP.
- The server code is described in excerpts; the full source is not published under an open licence.
- Nothing here shows SaaS companies paying a studio for this yet.
Sources
- Raphaël Aubry, How to expose your SaaS data to Claude and ChatGPT with MCP (DEV, Nov 6, 2026)
- Howseen MCP page (checked Nov 7, 2026)
- Official MCP registry entry ai.howseen/howseen-mcp (checked Nov 7, 2026)
- Sabba Keynejad (VEED CEO) on X: agents are becoming your users (Nov 6, 2026)
Sources reopened Nov 7, 2026. Spot a mistake? Tell us and we'll correct it in public.
Keep reading
- Cursor: plans and what people build and sell with it (guide)
- How much of the work do AI agents actually run? (guide)
- He still can't write code. His five-person ad agency now runs on about 270 tools he built with Claude. Every owner-run firm needs its first five
- A marketer who doesn't code built an elk-call coaching app with Claude Code and says it hit a $650K run rate in 110 days. Every skill learned by sound is the same opening
- A free n8n workflow flags fake supplier invoices before anyone pays them — and Germany's e-invoice switch makes it a service bookkeepers can sell
- Marc Lou's agent chased a broken-shoe refund through three support desks and got his 850 AED back. Small stores now need a claims desk that can answer agents
- Every discovery and launch plan