How to Use FORMLOVA with Gemini CLI -- One `gemini mcp add` Command to Create, Publish, and Run Forms From the Terminal

Last updated: June 15, 2026

Gemini CLI has become the tool you reach for in the terminal to research code, draft, and automate the odd task. But out of the box, that flow stops short of "build a feedback form, set up an auto-reply, and wire up a Slack notification." The moment you add an MCP server with a single command, that boundary disappears.

FORMLOVA exposes a remote MCP server that Gemini CLI can connect to with one gemini mcp add command. You can add it from the command line or by writing it straight into the mcpServers block of ~/.gemini/settings.json, and authentication is handled entirely by OAuth. Once connected, 130 tools across 25 categories line up at Gemini CLI's fingertips, so you can go from a draft to a preview, a pre-publish review, filtered responses, auto-replies, and workflow setup all in the same terminal conversation. This guide walks you from the connection through post-publish operations, the way to think about using write tools safely, and where to look when something breaks.

The Short Version -- One gemini mcp add Command and FORMLOVA Becomes Your Terminal Partner

There are only three things to do first.

1. In your terminal, run gemini mcp add --transport http formlova https://formlova.com/api/mcp. Leave it at the default to tie it to the project, or add -s user to make it a tool you carry across your machine.
2. Open Gemini CLI and run /mcp auth formlova. A browser opens, so sign in to FORMLOVA and click Authorize on the OAuth approval screen.
3. In your Gemini CLI session, say "Create a webinar signup form." A private draft and a preview URL come back.

That is it. There is no Bearer token to paste into settings.json by hand and no fixed secret to manage, because Gemini CLI supports remote HTTP MCP servers and authentication is completed by OAuth dynamic discovery. The whole thing takes roughly five minutes.

One note: Gemini CLI is actively developed as open source, and the names of command flags and config keys may change going forward. If the command names in this article ever differ from the actual behavior, follow Gemini CLI's latest documentation.

What Adding an MCP Server in Gemini CLI Actually Is

MCP stands for Model Context Protocol, an open standard that lets AI clients such as Gemini CLI access data and actions in external systems. Gemini CLI is Google's AI agent that runs in the terminal, and its MCP server management is concentrated in the gemini mcp subcommand and the mcpServers block of ~/.gemini/settings.json.

Translated into the world of form services, this means Gemini CLI reaches FORMLOVA's surface directly. No opening the dashboard in another tab, no downloading a CSV. Ask "Show me the unanswered responses for this form" and the list comes back on the spot.

The fact that everything stays in the terminal pays off less in the speed of building a form and more in the small operations that stack up after publication. While you write code or follow logs, in that same terminal you sort responses by temperature, tailor an auto-reply to the recipient, move a reminder up a day, or filter out sales-pitch responses before you analyze. The thinking behind this is covered in detail in Most form tools stop at creation.

The Connection Flow -- Add It in One Shot With gemini mcp add

Let's walk through it in three steps. The operation is the same on macOS, Linux, and Windows.

1. Confirm the FORMLOVA MCP Server URL

The FORMLOVA MCP server URL is a single value, shared across every client.

https://formlova.com/api/mcp

No trailing slash. It is not an older-style path like /api/sse either. Whether you connect from Gemini CLI, Claude Desktop, Claude Code, ChatGPT, or Cursor, you use this URL. If you do not have a FORMLOVA account yet, sign up on the free plan first from the FORMLOVA setup guide, because you will be asked to sign in via OAuth when you authenticate.

2. Run the gemini mcp add Command

Run the following command in your terminal. The --transport http flag tells Gemini CLI that FORMLOVA is a remote, HTTP-streaming MCP server.

gemini mcp add --transport http formlova https://formlova.com/api/mcp

Breaking down the syntax: formlova is the server name, https://formlova.com/api/mcp is the connection URL, and --transport http makes the HTTP transport explicit. If you do not specify a scope, the entry is written by default to the project-level .gemini/settings.json. To use it across every project you touch, add -s user (--scope user) so it is written to ~/.gemini/settings.json in your home directory.

gemini mcp add -s user --transport http formlova https://formlova.com/api/mcp

You can reach the same result by editing the config file directly instead of using the command. Just add the following to the mcpServers block of ~/.gemini/settings.json. A remote HTTP server is specified with the httpUrl key.

{
  "mcpServers": {
    "formlova": {
      "httpUrl": "https://formlova.com/api/mcp"
    }
  }
}

Here, httpUrl is the key for an HTTP-streaming connection destination, and it is different from command, which points to a local executable, and from url, which points to an SSE-style destination. Because FORMLOVA is a remote HTTP server, you use httpUrl. Adding it with the command writes the same content to this file, so the result is identical. If you write it by hand, just watch out for JSON syntax mistakes (missing commas or unclosed brackets).

3. Complete OAuth Authorization and Check the Tool List

Once the config is written, run /mcp auth formlova in your Gemini CLI session. The browser opens automatically, switches to the FORMLOVA sign-in screen, and shows the OAuth approval screen. Press Authorize and the browser closes, and back in the terminal Gemini CLI can now use FORMLOVA's tools. The FORMLOVA MCP server supports OAuth configuration via dynamic discovery, so there is no need to prepare a client ID or secret by hand.

Run /mcp to see the registered MCP servers, their connection status, and the list of tools each provides. If formlova shows as connected, you are done. 130 tools line up across 25 categories -- form creation, response management, email, analytics, workflows, and more -- but the first time you only use two or three. There is no need to feel overwhelmed by how many there are.

Choosing a Scope -- project / user

Gemini CLI lets you choose where the MCP config is written. By default (with no scope specified), the entry is written to the .gemini/settings.json at the root of the project where you ran the command. Include that file in Git and every teammate can share the same MCP server config. This is the right fit when you want to share a product's operational forms across the team so that anyone who launches gemini can edit them.

Specify -s user (--scope user) and the entry is written to ~/.gemini/settings.json in your home directory, making FORMLOVA available across every project you touch. This is the easier choice for usage that is not tied to a specific project -- managing a personal contact form, a blog reader survey, or a quick research form.

If you run FORMLOVA with a team, committing the project-scope .gemini/settings.json to Git is the easier path. The moment a new member clones the project, the same form operations environment comes up. The rule of thumb -- user scope when you use it across every project you touch -- should keep you out of trouble.

Note that the OAuth credentials themselves are stored per user. Even if you share the config under project scope, each member authenticates with their own account, so permission separation is preserved.

Your First Conversation -- Build a Form Draft From the Terminal

Once you are connected, type a short line into your Gemini CLI session. FORMLOVA is designed so that you do not need to write a long prompt to start.

Create a contact form.

That is enough. Gemini calls FORMLOVA's form-creation tool, assembles fields that fit the purpose, and returns a private draft. The reply looks like this:

I created a draft of your contact form.

Fields:
- Name (required)
- Email (required)
- Company (optional)
- Inquiry category (required, select)
- Message (required)
- Consent to the use of personal information (required)

Preview URL: https://formlova.com/preview/...

Click the preview URL and the actual form screen opens in a new tab. You can check the title, field order, required and optional badges, the consent checkbox, and the submit button from the respondent's point of view. Being able to refine while watching the preview in the window next to the terminal is the good part about pairing this with Gemini CLI.

From here, you refine through chat while looking at the screen. Short instructions are fine.

Add a phone number field below company and make it optional.
Add "About hiring" to the inquiry category.
Put our privacy policy URL into the personal-information consent text.

Reopen the same preview URL and it refreshes to the latest state. Look at the real screen and say only what is missing. This back-and-forth is the main reason to pair FORMLOVA with Gemini CLI. The general creation flow is also summarized in How to create forms with ChatGPT and Claude.

Check the Preview and Run the Pre-Publish Review

Once the draft takes shape, you move to the pre-publish review. In FORMLOVA, asking "please publish" does not publish immediately. A checklist runs on the server side, and if there are unset blocking items, they come back along with the review result.

Publish this form.

The reply looks like this:

There are items to confirm before publishing.

- Preview URL: https://formlova.com/preview/...
- Thank-you page preview: https://formlova.com/preview/.../thankyou
- Duplicate submission prevention: not set
- Privacy policy URL: not set
- Sales-email auto-detection: not confirmed (text input fields present)

Open the preview and the thank-you page preview in separate tabs and visually check the copy, field order, and required settings. When you are done, return to the terminal and reply to Gemini like this:

I checked both the preview and the thank-you page. Prevent duplicate
submissions by email. Use https://example.com/privacy for the privacy
policy. Enable sales-email detection. Publish it now.

Gemini obtains the token for final publication, publishes the form, and the public URL comes back. A guide that follows this pre-publish review step by step is at How to run the pre-publish review and publish with FORMLOVA.

Run Post-Publish Operations From the Terminal

The real work begins once responses start coming in. This post-publish part is exactly where the Gemini-CLI-and-FORMLOVA combination earns its keep. Here are three representative moves.

Filter the Response List to Check the Situation

As responses grow, filtering by condition is faster than scanning everything. Without stopping the work in front of you, ask Gemini in the same terminal:

From this week's inquiries, show me only the unanswered ones.
Pull out only the responses categorized under "About pricing."
Exclude the sales-pitch responses and list only the genuine inquiries.

Only the matching responses come back as a table, and you can drill into a response of interest, change its status to in progress or resolved, and add a note -- all in the same conversation. Details are summarized in How to view, filter, and manage form responses with FORMLOVA.

Adjust the Auto-Reply Email Copy and All

On the standard plan and higher, you can freely customize the auto-reply email. Ask Gemini and it can draft the copy and even run a test send.

Make this form's auto-reply email a bit softer in tone.
Open with "Thank you for your inquiry."
Put our business hours (weekdays 9 a.m.-6 p.m.) at the end.
When it's ready, send a test email to my address.

Gemini presents the new copy as a preview and asks "Save this version?" It always inserts a confirmation before saving, so the reply on a live form is never rewritten on its own.

Add Automation With a Workflow Place Recipe

For automation like "notify Slack when a response comes in" or "send a follow-up email three days later," copying a recipe from Workflow Place is the shortcut. Gemini shows candidates, and the prompt for the recipe you choose can be handed straight to execution. See How to find a recipe in FORMLOVA's Workflow Place and set it up from chat for details.

Using Write Tools Safely -- The Trust-Boundary Mindset

An MCP server in Gemini CLI can call not only read operations but also write operations from the conversation -- publishing a form, updating a response, sending email. Convenient as that is, it is a mechanism that grants an external server the right to act, so a few points keep it safe.

First, only add MCP servers from trustworthy sources. For FORMLOVA, register only the official formlova.com URL. Adding a URL of unknown origin with gemini mcp add or writing it into settings.json means you are permitting that server's behavior through Gemini CLI.

Second, FORMLOVA provides a safety valve of its own. High-impact operations such as publishing, deleting, and sending email require a confirmation on the server side. Even if you ask "please publish," the server returns a confirmation checklist and will not publish until explicit approval and a token are both in place. This is guaranteed by the FORMLOVA server, not by Gemini CLI.

Finally, FORMLOVA is designed to treat the responses of external services and the contents of public forms as data, and not to follow instruction-like text mixed into them. In other words, a command slipped into a response body or imported external data will not trigger a publish or send on its own. Precisely because write tools can be called from a conversation, the final trigger stays with human approval.

Where to Look When Things Go Wrong

There are failure modes specific to a CLI, so here are the few places worth checking.

When the gemini command is not found, Gemini CLI itself may not be installed, or it is not on your PATH. Check with which gemini, and if it returns nothing, revisit Google's official Gemini CLI installation instructions.

When gemini mcp add succeeded but it does not show in /mcp, the scope it was written to may differ from what you expected -- a case where you meant to pass -s user but it actually went to project scope. Look inside both the project-level .gemini/settings.json and the home-directory ~/.gemini/settings.json to confirm which mcpServers it was written to. If you wrote it by hand, also check that the JSON syntax (commas and brackets) is not broken.

When you cannot connect, check the URL first. See whether you entered anything other than https://formlova.com/api/mcp, and whether a trailing slash or an older-style path like /api/sse is still attached. Also confirm you specified it with the httpUrl key (and did not mix it up with url or command). Because FORMLOVA is a remote HTTP server, you use httpUrl.

When an "Unauthorized" error comes back, it is a sign your OAuth authorization has expired. Run /mcp auth formlova again to re-authenticate.

When the authorization screen never appears, a browser popup blocker is almost always the cause. Allow popups for formlova.com and try again. If you run Gemini CLI in a remote environment over SSH, the browser may not open, so it is safest to authenticate once in a local session where a browser can open before you work.

When tools do not show, or only some appear, exit your Gemini CLI session, restart it, and check the state with /mcp. Loading the tool list can take a moment after the first connection.

When you speak Japanese but get English replies, your profile language is set to English. One line -- "Set my profile language to Japanese" -- switches it. A complete list is in the troubleshooting section of the FORMLOVA setup guide.

What You Can Do by Plan

Connecting from Gemini CLI itself works on every FORMLOVA plan. The difference shows up in the range of features available after you connect. Prices include tax.

Even on the free plan, you can experience the whole loop from draft to publish, response management, and CSV export. When you want to write the auto-reply in your own voice or send reminders automatically, move to standard. When bulk sends or paid events enter the picture, move to premium. Trying them in order and upgrading only for the feature you actually need is plenty.

Related Workflows You Can Use as-is

Once you have built a form in Gemini CLI and run it through the publish review, Slack Notification + Sheets Log is the easiest way to move into operations. You can flow the responses from a form you built in the terminal straight into team notifications and a record.

To connect into sales or knowledge management, combine Response to HubSpot Contact or Response to Notion Database. Rather than stopping at "I built a form in Gemini CLI," put the responses into a state where they move to their next destination.

FAQ

Which is correct, gemini mcp add or editing settings.json directly?

Both reach the same result. Running gemini mcp add --transport http formlova https://formlova.com/api/mcp writes the same entry into the mcpServers block of settings.json. The command is less prone to syntax mistakes, so it is the easier choice day to day, but if you want to embed it in a CI setup script or manage the contents in detail, editing ~/.gemini/settings.json directly is fine too.

Should I choose project scope or user scope?

For team development where you share and operate the same forms, use the default project scope (.gemini/settings.json); to use it as a tool of your own, use -s user (~/.gemini/settings.json). Commit the project-scope config file to Git and the moment a new member clones the project, they get the same environment. OAuth credentials are stored per user, so sharing the config never means shared permissions.

Do I specify a remote server with url or httpUrl?

Because FORMLOVA is a remote, HTTP-streaming MCP server, in mcpServers you set https://formlova.com/api/mcp on the httpUrl key. url points to an SSE-style destination, and command points to the command for an MCP server you run locally -- different purposes. If you add --transport http to gemini mcp add, it is written as httpUrl automatically.

Do I need to prepare a client ID or secret by hand?

No. The FORMLOVA MCP server supports OAuth configuration via dynamic discovery, so authentication is complete simply by running /mcp auth formlova and approving the sign-in in the browser. There is no need to paste a fixed token into settings.json.

Is the URL the same for other clients (Claude Desktop, Claude Code, Cursor, ChatGPT)?

Yes. The FORMLOVA MCP server URL is a single https://formlova.com/api/mcp, and you register the same URL from every client. If you want to try it casually in a GUI, see the complete guide to using FORMLOVA with Claude Desktop; for a different CLI, the complete guide to using FORMLOVA with Claude Code; for writing inside an IDE, the complete guide to using FORMLOVA with Cursor; from ChatGPT in the browser, the complete guide to using FORMLOVA with ChatGPT. What you can do is the same from any client.

I am worried Gemini CLI might publish or delete things without my approval.

High-impact operations like publishing, deleting, and sending email require a confirmation on the server side. Even if you ask "please publish," the server returns a confirmation checklist and will not publish until explicit approval and a token are both in place. This is guaranteed by the FORMLOVA server, not by Gemini CLI. Even in a design where write tools can be called from a conversation, the final trigger stays with human approval.

Related Articles

• The MCP Form Service Guide for FORMLOVA -- Not Just Building With AI, but Running Forms After Publication
• The Complete Guide to Using FORMLOVA with Claude Desktop -- From Adding the Connector to Publishing and Running Forms
• The Complete Guide to Using FORMLOVA with Claude Code -- From One CLI Command to Form Creation, Publishing, and Operations in the Terminal
• The Complete Guide to Using FORMLOVA with Cursor -- From In-IDE MCP Setup to Form Creation, Publishing, and Operations
• The Complete Guide to Using FORMLOVA with ChatGPT -- From a Developer-Mode Custom Connector to Publishing and Running Forms
• How to Create Forms with ChatGPT and Claude -- With FORMLOVA, Go From One Line to a Draft and Preview
• How to Run the Pre-Publish Review and Publish with FORMLOVA
• How to Find a Recipe in FORMLOVA's Workflow Place and Set It Up From Chat

Try FORMLOVA for free

Official Sources Referenced

• Google: Gemini CLI -- MCP servers with the Gemini CLI (accessed 2026-06-15)
• Model Context Protocol: Introduction (accessed 2026-06-15)
• FORMLOVA setup guide (accessed 2026-06-15)

Disclosure and Verification

This guide is written by the FORMLOVA development team. The Gemini CLI connection steps were reflected into this article after verifying Google's official Gemini CLI MCP server documentation, the Model Context Protocol specification, and the implementation of the FORMLOVA setup guide on June 15, 2026. Gemini CLI is actively developed as open source, and the flags of the gemini mcp subcommand, the config keys in settings.json (mcpServers / httpUrl / authProviderType and the like), and the handling of OAuth can change on Google's side. Before adopting this, confirm Gemini CLI's latest official documentation.