BananaBanana MCP Server

BananaBanana MCP is a hosted remote MCP server for generating and editing AI images, video, and speech directly from MCP-compatible clients and AI agents.

It provides a single MCP endpoint for working with multiple generation models from providers including Google, OpenAI, Alibaba, and xAI, while using one shared pay-as-you-go BananaBanana balance.

No monthly subscription is required.

What you can do

With BananaBanana MCP, an AI agent can:

• Generate images from text prompts
• Generate images using reference images
• Edit existing images
• Generate video from text
• Generate video from an image
• Edit and extend generated video
• Generate speech from text
• Create single-speaker and supported multi-speaker speech
• Discover currently available generation models
• Check live model pricing before starting a generation
• Check account balance and account information
• Top up the BananaBanana balance
• Track generation jobs and retrieve completed results
• Browse recent generation history

The same BananaBanana account, balance, and generation history can be used across both the web application and MCP integrations.

---

Supported AI models

BananaBanana provides access to multiple image, video, and speech model families through one MCP interface.

Supported model families include:

Image

• Google Nano Banana / Gemini image models
• OpenAI GPT Image
• Alibaba Qwen Image

Depending on the selected model, workflows can include:

• Text-to-image
• Reference-based image generation
• Image editing
• Multi-image workflows
• Different resolutions and aspect ratios

Video

• Google Veo
• Gemini Omni video workflows
• Alibaba Wan
• xAI Grok Imagine Video

Supported workflows include:

• Text-to-video
• Image-to-video
• Video editing
• Video extension
• Reference-driven generation
• Synchronized audio on supported models

Speech

• Gemini TTS and supported speech-generation models

Speech generation can be used for narration, dialogue, voice content, and other agent-driven audio workflows.

Available models and pricing can change over time, so MCP clients can use list_models to retrieve the current model catalogue and live pricing instead of relying on hard-coded model information.

---

MCP endpoint

BananaBanana runs as a hosted remote MCP server.

https://bananabanana.pro/api/mcp

There is no need to download or run the generation server locally.

Transport

BananaBanana uses Streamable HTTP for remote MCP connections.

---

Authentication

BananaBanana supports two authentication approaches.

OAuth 2.1

OAuth is the recommended connection method for MCP clients that support remote MCP authentication.

Add the BananaBanana MCP endpoint:

https://bananabanana.pro/api/mcp

The client can then guide the user through the BananaBanana authorization flow.

This avoids manually copying API keys and is the preferred setup for interactive MCP clients.

Bearer API key

API keys are also available for clients, scripts, development environments, CI systems, and integrations where OAuth is not practical.

Example authorization header:

Authorization: Bearer bb_live_YOUR_KEY

An API-key based connection can also be used through tools such as mcp-remote.

Example:

{
  "mcpServers": {
    "bananabanana": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://bananabanana.pro/api/mcp",
        "--header",
        "Authorization: Bearer bb_live_YOUR_KEY"
      ]
    }
  }
}

---

Available MCP tools

BananaBanana currently exposes 10 MCP tools.

list_models

Returns the currently available generation models and their capabilities.

Use it to discover:

• Available image models
• Available video models
• Available speech models
• Supported workflows
• Current pricing
• Model-specific parameters

This allows agents to choose a model dynamically instead of relying on a fixed list.

get_account

Returns BananaBanana account information such as the current balance.

Agents can use this before expensive generation tasks to verify that sufficient balance is available.

top_up

Provides balance top-up functionality from an MCP workflow.

This makes it possible for an agent to guide the user through adding funds when additional generation balance is required.

generate_image

Generates a new AI image.

Depending on the selected model, it can be used for:

• Text-to-image
• Reference-image generation
• Multi-reference workflows
• Different aspect ratios
• Different output resolutions

edit_image

Edits an existing image using an AI model and natural-language instructions.

Typical workflows include:

• Changing objects or visual details
• Restyling an image
• Editing a subject while preserving the rest of the image
• Reference-guided modifications
• Product and creative image transformations

generate_video

Generates AI video from text or image inputs.

Depending on the selected model, supported workflows can include:

• Text-to-video
• Image-to-video
• Reference-based video generation
• Different durations and resolutions
• Synchronized audio

edit_video

Edits or extends supported videos using compatible video models.

This can be used for workflows such as:

• Video continuation
• Video extension
• Prompt-based video modification
• Supported reference-driven transformations

generate_speech

Generates speech from text.

It can be used for:

• Narration
• Voice-over
• Dialogue
• Agent-generated audio
• Supported multi-speaker workflows

get_result

Checks the status of an asynchronous generation job and returns its result when processing has completed.

This is particularly useful for video generation, where jobs may take longer than a normal MCP request.

Typical agent workflow:

generate_video
      ↓
generation ID
      ↓
get_result
      ↓
processing
      ↓
get_result
      ↓
completed result

list_generations

Returns recent generation history associated with the BananaBanana account.

This allows agents to inspect previous jobs and retrieve information about past generations.

---

Pay-as-you-go pricing

BananaBanana uses a prepaid pay-as-you-go balance rather than a monthly subscription.

You only pay for generations that you run.

There are:

• No required monthly subscriptions
• No seat-based plans
• No requirement to maintain a recurring plan
• No separate MCP subscription

Generation prices depend on the selected model and generation parameters.

Current entry prices include:

• Images from approximately $0.03
• Video from approximately $0.09
• Speech from $0.01 per started 200 transcript characters

Because model pricing can change, agents should use list_models to retrieve current pricing before generation.

---

Cost-aware agent workflows

BananaBanana MCP is designed so AI agents can reason about generation cost before committing to expensive operations.

A typical autonomous workflow can look like this:

User requests a video
        ↓
Agent calls list_models
        ↓
Agent compares available models and prices
        ↓
Agent selects an appropriate model
        ↓
Generation request is prepared
        ↓
User can see the cost before an expensive operation
        ↓
Generation starts
        ↓
Agent checks get_result
        ↓
Final media is returned

This is particularly useful for video generation, where the difference in price between models, duration, quality levels, and resolutions can be significant.

---

Payments and balance

BananaBanana uses one shared prepaid balance for web and MCP generations.

Supported funding methods include:

• Cryptocurrency
• Card payments
• PayPal
• SEPA

Crypto top-ups can start from a relatively small amount, while card, PayPal, and SEPA payments have a higher minimum top-up.

The MCP server can check the account balance before generation, allowing an agent to detect insufficient funds before attempting an expensive job.

---

Generation safety and billing behavior

Generation operations are designed around explicit job and pricing information.

For expensive operations, clients can obtain pricing information before generation rather than blindly triggering a costly model.

If a generation fails under supported failure conditions, the corresponding generation charge is refunded automatically.

This makes BananaBanana suitable not only for interactive use but also for automated agent workflows where failed generations should not silently consume the user's balance.

---

Example agent workflow

A user could ask an MCP-enabled assistant:

Create a cinematic image of a futuristic city, then animate it into a 5-second video.

The agent can perform the workflow through BananaBanana:

1. list_models
   ↓
2. Choose an image model
   ↓
3. generate_image
   ↓
4. get_result
   ↓
5. Choose an image-to-video model
   ↓
6. generate_video
   ↓
7. get_result
   ↓
8. Return the final video

The agent can inspect available models and prices at each stage instead of having model names and pricing hard-coded into the client.

---

Why use BananaBanana through MCP?

One API for multiple model providers

Instead of integrating several separate image, video, and speech APIs, agents can access multiple model families through one MCP interface.

Image + video + speech in one workflow

A single agent can move from:

Prompt
→ Image
→ Edited image
→ Video
→ Speech

without switching between unrelated MCP servers.

Built for remote MCP

BananaBanana is hosted, so the user does not need to run GPU infrastructure, model servers, or a local BananaBanana backend.

Live model discovery

Agents can retrieve currently available models and pricing dynamically through list_models.

Shared balance

Web and MCP generations use the same BananaBanana account and prepaid balance.

No subscription

Occasional users and automated agents can pay for the generations they actually use rather than maintaining another monthly AI subscription.

---

Use cases

BananaBanana MCP can be used for:

• AI-agent image generation
• Automated social media content
• AI video creation
• Image-to-video workflows
• Product visualization
• Ad creative generation
• Storyboards
• Short-form video
• AI-generated narration
• Automated creative pipelines
• Marketing content
• Developer tools
• MCP-enabled IDE workflows
• Autonomous content agents
• Multi-step image → video → speech workflows

---

Links

Website

https://bananabanana.pro/

MCP documentation

https://bananabanana.pro/mcp

Remote MCP endpoint

https://bananabanana.pro/api/mcp

GitHub

https://github.com/bananabanana-pro-mcp/bananabanana-mcp

Official MCP Registry package

pro.bananabanana/image-video

---

Quick start

For an MCP client with native remote-server and OAuth support, add:

https://bananabanana.pro/api/mcp

and authenticate with your BananaBanana account.

For API-key based integrations, use the same endpoint with:

Authorization: Bearer bb_live_YOUR_KEY

Once connected, call:

list_models

to discover the latest available image, video, and speech models and their current prices before starting your first generation.