Gemini · BridgeheadBridgehead
NewGemini 4 Argon and Nano Banana 2.1 are here

Gemini API × every MCP app

All of Gemini, in every app you work in.
On your own Gemini key.

Images · Video · Vision · Answers · Search · Research · Consult · Speech · Music · Transcripts

View DemoUse Cases

Google's newest. Usable here today.

New

Gemini 4 Argon

Google's new flagship model, for answers and hard reasoning. Released to developers on 13 October.

Say it by name

Use Gemini 4 Argon to explain this contract in plain words.
New

Nano Banana 2.1

Google's newest image model, released on 6 October.

Say it by name

Make a picture of a lighthouse at dusk with Nano Banana 2.1.
New

Fill a form from another

Carry your answers from last year's PDF form onto this year's, even when the form has been redesigned. Answers it's sure of are filled in; the rest come back for you to check.

Try saying

Fill this year's W-9 from my old one.

There is no menu to pick from. Name the model in your chat and the bridge passes it straight to Google, on your own Gemini key.

Google's Gemini API, with Nano Banana for images and Omni for video, works on an API key, and claude.ai, ChatGPT, the phone and voice have no place to put one. Gemini · Bridge is a connector you add once: sign in, your Gemini API key from Google AI Studio goes into Bridgehead's vault, and Nano Banana, Omni video, Gemini's answers, Deep Research, speech, Lyria music and transcripts are there in every app you work in, and any AI there can consult Gemini or hand it work. Switch each one on or off on your key page; everything you make is kept in your Bridgehead gallery.

THE USUAL WAY NO PLACE FOR AN API KEY GEMINI · BRIDGE Claude / Claude CodeChatGPT / Codex Gemini Claude / Claude CodeChatGPT / Codex GEMINI · BRIDGE Gemini SIGN IN ONCE · ALL OF GEMINI IN ANY APP

The full demo. 1 min 56 s.

The whole film · with sound

What it can do

Images · Nano BananaMake and edit pictures: from your photos, keeping characters the same or matching a style, from a video or a YouTube link, grounded in Google Search; PNG or JPEG.
  • From a prompt, or edit your own photos.
  • Keep people or characters looking the same, and match the style of other pictures.
  • From a video or a public YouTube link, such as a thumbnail or a poster.
  • Change a picture step by step by talking; each change builds on the last.
  • Grounded in Google Search, with its sources.
  • PNG or JPEG.
Video · OmniVideos with sound on Gemini Omni Flash: from a prompt, a picture or a first and last frame, then changed by talking and continued.
  • Sound, speech and music, steered in the prompt; 360p to 4K.
  • From a prompt, a picture, a first and last frame, or up to 3 reference images.
  • Change a video by talking ("make it snow"); edits build on each other.
  • Continue a video a few seconds at a time.
  • Veo 3.1 too, until Google retires it on 22 Oct 2026.
Answers · GeminiAsk about files up to 500 MB, YouTube videos or one part of a recording; it can search Google and give its sources.
  • Documents, pictures, recordings and videos, up to 500 MB each.
  • Last year's answers on this year's form.
  • Public YouTube links, which Google watches itself.
  • Just one part of a video, from a start time to an end time.
  • Google Search, with the sources it used.
Consult and delegate · GeminiAsk Gemini for a second opinion, a critique or a comparison, hand it a subtask, or go back and forth with it; and raw access to any Gemini text model.
  • A second opinion, a critique or a comparison of drafts, from a different model.
  • Hand Gemini a subtask and get back only the result, so your main assistant keeps its own context.
  • Back and forth: go on with the same conversation by its id.
  • Its strengths: files up to 500 MB, YouTube, Google Search with sources, and web pages you name.
  • A thinking level, and answers in a JSON shape you give.
  • Raw access: any Gemini text model with all its settings, for whatever else you think of.
SpeechRead text aloud in hundreds of voices from Google's library, or a voice you design from a description; one speaker or two.
  • Google's voice library, by language, accent, gender and persona.
  • A voice designed from a description, kept a year from its last use.
  • Two speakers in conversation, with the 30 prebuilt voices.
Music · LyriaA 30-second clip or a whole song with its lyrics, from a prompt or pictures.
TranscriptionA recording or video into text, with times and speakers, or translated.
  • Gemini 3.5 Transcribe: each word's time, up to 1,000 names and terms, about half a cent a minute.
  • Times and speaker labels, or plain text.
  • Translated into a language you name.
  • Recordings, videos and public YouTube links.
Deep ResearchA cited research report on any question: Google's agent searches and reads for up to an hour; about $1–3 a run.
  • A report with its sources, and charts when they help.
  • Your own pages and PDFs to read too.
  • Max for the hardest questions, about $3–7.
  • Ask about a finished report without a new run.

Copy. Paste. Connected.

How to connect it

01
Copy the addressgeminibridge.fyi/mcp
02
Add it to your appClaude, ChatGPT, Codex or any MCP app
03
Sign inOne account, made the first time
04
Add your Gemini API keyOnce. Checked, then kept encrypted
05
AskEvery Gemini model, wherever you work

Stuck or found a bug? Tell us

MCP server

The address to paste into each app, or its one-line command. It's the same connector everywhere.

Claude

claude.ai, the desktop app or the phone: Customize → Connectors → Add → Custom connector, then paste this.

https://geminibridge.fyi/mcp

Claude Code

One line in your terminal.

claude mcp add --transport http geminibridge https://geminibridge.fyi/mcp

ChatGPT

In settings, turn on developer mode and add a custom app with this address.

https://geminibridge.fyi/mcp

Codex

One line in your terminal; it asks you to sign in.

codex mcp add geminibridge --url https://geminibridge.fyi/mcp

Any other MCP app

Cursor, VS Code, Windsurf, Zed, n8n and more: add this address as a remote MCP server and sign in when asked.

https://geminibridge.fyi/mcp

Your Gemini API key

Sign in when your chat asks, then add your Gemini API key from Google AI Studio once. You pay Google directly.

Add your Gemini key

Questions

What can I do with Gemini · Bridge?

Make and edit pictures with Nano Banana: from a prompt, from your own photos, keeping characters the same or matching a style, from a video or a YouTube link, grounded in Google Search for real places and products, and changed step by step by talking. Make videos with sound on Gemini Omni, from a prompt, a picture or a first and last frame, then change them by talking and continue them. Ask Gemini about files up to 500 MB, YouTube videos, or one part of a recording; have Deep Research write a cited report; turn text into speech in hundreds of voices, or one you design; make music with Lyria; and get transcripts with each word's time and speakers. Consult Gemini from any app or agent: a second opinion, a critique, a comparison, or a subtask handed over and its result handed back, going back and forth in one conversation; or use any Gemini text model raw, with all its settings. Everything it makes lands in your gallery, and you can switch any of it on or off on your connector's page.

Can another AI hand work to Gemini?

Yes. Claude, ChatGPT, Codex or any agent that speaks MCP can consult Gemini as a second model: get its take on an answer, have it critique or compare drafts, or hand it a subtask (reading a 500 MB file, a YouTube video, a search with sources, a web page) and fold the result back into its own answer, without spending its own context. Choose a thinking level, ask for a JSON answer in your own shape, and carry on the same conversation by its id. For anything else, the raw tool takes any Gemini text model with its own settings: temperature, stop words, a JSON schema, thinking, code execution.

Which Gemini models can I use?

Nano Banana 2, Nano Banana Pro, Nano Banana 2 Lite and the newest, Nano Banana 2.1, for images; Gemini Omni Flash for video (and Veo 3.1 until Google retires it on 22 Oct 2026); Gemini Flash, Pro and Flash-Lite for answers, which can read your pictures, PDFs, audio and video and search Google, and any other Gemini text model by its name, Gemini 4 Argon among them; Deep Research and Deep Research Max for reports; Gemini's speech models with Google's voice library; Lyria 3 for music, a clip or a whole song; and Gemini 3.5 Transcribe for transcripts. Turn any of them off on your key page. The model list comes from your own key.

Can it fill in a PDF form for me?

Yes, from one you've already filled, even a scan or phone photos of it. Give it the old form and the new one: this year's version, or a different form that asks some of the same questions. It matches questions by their printed words, so it still works when the field names change. Answers it's sure of go straight in. Anything less certain, like an address, which can change, comes back for you to confirm. A scan or photo is read by Gemini on your own paid Gemini key, and an answer read off the page goes in only when it reads clearly. For a typed form, Gemini sees only the printed questions, never your answers. Your answers aren't kept. It gives no tax or legal advice.

Is my Gemini API key safe?

It's checked, kept encrypted in Bridgehead's vault and used only to call Google for you. You pay Google directly; remove the key at any time at geminibridge.fyi/key.

Where do my images and videos go?

Into your Bridgehead gallery, each with a link you can share. Google deletes what it makes after a couple of days, so Bridgehead keeps its own copy.

What does Deep Research cost?

It runs on your own Gemini key: about $1–3 a report, or $3–7 with Deep Research Max, depending on how much it searches and reads. Google bills you directly. A question about a finished report is quick and costs much less.