AI-native form backend
Your agent sets up the whole form backend.
A form backend whose MCP server does everything the dashboard does — 29 tools. Watch an agent set one up without opening a browser once.
Claude Code v2.1.270
Opus 5 (1M context) · Claude Max
~/Code/acme
https://shipmyform.com/api/mcpNo step sends you to a browser
Most MCP integrations stop at “create a resource” and hand you back a dashboard link. This one covers the lifecycle.
- 1
It creates the account
The agent asks for your email and calls sign_up. No signup form, no password, no browser tab. A 6-digit code goes to your inbox.
- 2
You approve it with six digits
You read the code and hand it over. The email says in plain English what the agent's key will be able to do before you approve it.
- 3
It builds the whole thing
Forms, validation rules, spam settings, Slack or Sheets or webhook delivery, a hosted page if the site isn't ready. All of it, as tools.
- 4
It runs the inbox
List, search, label and export submissions from wherever you already work. Labelling teaches the spam classifier, so triage compounds.
29 tools, not a token integration
| Area | Tools | Covers |
|---|---|---|
| Account | 9 | sign_up, verify_account, API keys, plan and billing |
| Forms | 6 | create, read, update, delete, snippets, stats |
| Connectors | 6 | list types, add, test, enable, disable, remove |
| Hosted pages | 2 | publish and unpublish a standalone page |
| Submissions | 5 | list, read, label, delete, export CSV |
Every tool, with arguments: the MCP docs
Scoped access
The agent that builds your site can't read your inbox
Every key carries a scope set, enforced on the MCP server and the REST API alike. A key can never grant scopes it does not itself hold, so an agent cannot widen its own access — handing it a narrow key is a real limit, not a suggestion.
| Preset | What it can do | Give it to |
|---|---|---|
| wiring | Builds forms. Cannot read a single submission. | The agent scaffolding your site. |
| triage | Reads and labels submissions, never deletes. | An inbox assistant. |
| readonly | Looks at everything, changes nothing. | Reporting and analysis. |
| full | Everything, including deletion and billing. | Your own trusted agent. |
What still needs a person
Two things, deliberately. The verification code, because a form backend that an agent can create for someone else's address would be a spam tool. And payment: start_upgradereturns a checkout URL, but a card is typed by a human on the provider's page. Everything between those two points is the agent's.
Questions, answered
Depth. Most expose a handful of tools — create a form, read some stats — and send you to the dashboard for everything else. Here the MCP server is the whole product: 29 tools covering the account lifecycle, form configuration, connectors, hosted pages, billing and the inbox. An agent never needs to hand you back to a browser.
By default it asks you for it, and the server's own instructions tell it to. If you have connected an email tool to your agent it can read the code itself, which is convenient when it is your mailbox and your agent. The code proves control of the address, which is the same standard as a magic link.
You don't have to. Keys are scoped, and the wiring preset cannot read submission content at all — so the agent that builds your site, and reads untrusted pages while doing it, has no path to your customers' messages. Scopes are enforced on the MCP server and the REST API alike, and a key can never grant scopes it does not itself hold.
Anything that speaks MCP over Streamable HTTP: Claude Code, Claude Desktop, Cursor, VS Code, Codex, Gemini CLI, Windsurf and others. The endpoint is https://shipmyform.com/api/mcp and it authenticates with a Bearer API key. Per-client setup is in the docs.
Not for the wiring. Creating the account, forms, snippets, connectors, hosted pages and stats all work on the free plan, 250 submissions a month, no card. Reading, classifying and exporting submission content through MCP or the REST API is a paid feature; your submissions are always visible in the dashboard either way.
Yes, with instant endpoints: point a form at /to/your@email and it works from the first submission. That is the zero-friction path when the agent has nothing but an address. The MCP server is the richer one, because it gets a real form ID and keeps your address out of the page.
Point your agent at it
Free plan, 250 submissions a month, no card. Your agent can do the signup too — it only needs your email address.