nano-banana-gemini
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| FLOW_LANG | No | Language of this server's messages: `en` or `es`. Defaults to the system locale, else `en`. | |
| FLOW_CDP_URL | No | Force a Chrome debugging endpoint (`http://…` or `ws://…`). Unset: your everyday Chrome if remote debugging is allowed, else `http://127.0.0.1:9222`. | |
| FLOW_MAX_COST | No | Credit ceiling per generation. | 0 |
| FLOW_OUTPUT_DIR | No | Where images are saved. | ~/nano-banana-images |
| FLOW_PROFILE_DIR | No | Browser profile folder. | ~/.nano-banana-mcp/chromium-profile |
| FLOW_BROWSER_MODE | No | `headless` (default) or `hidden`. Use `hidden` if Google blocks headless. | headless |
| FLOW_CHROMIUM_PATH | Yes | Path to the Chromium binary. Setting it turns on the background browser. Required for this fork's setup. | |
| FLOW_AUTOLAUNCH_PORT | No | Debug port used by the background browser. | 9333 |
| FLOW_CHROME_USER_DATA | No | Where to look for `DevToolsActivePort`, if your Chrome uses a non-default user-data folder. Defaults to Chrome's default folder. | |
| FLOW_GENERATE_TIMEOUT_MS | No | Max wait per image. | 180000 |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| flow_statusA | Check the connection to the Google Flow Chrome window: signed-in session, account, open project and credit balance. Call this first when something fails. |
| generate_imageA | Generate one or more images with Google Flow and save them to disk. You can ask for an exact pixel size (e.g. 1200x630): it generates at the closest native aspect ratio and crops to the size you asked for. Images cost no credits. |
| download_imageA | Download an image that already exists in Flow by its media id, with optional cropping to an exact size. |
| upscale_imageB | Upscale an image that already exists in the open Flow project, by its media id, and save it to disk. 2K is free. |
| list_libraryA | List the images in the Flow project library: uploads (reusable via reference_library_names without re-uploading) and generated images (re-downloadable by id with download_image). |
| generate_batchA | Generate several DIFFERENT prompts in one call. All of them go to the SAME Flow tab, queued: each prompt is sent while the previous ones are still generating, and everything is collected at the end. No extra tabs are opened. Up to 8 jobs; each can have its own size, filename and reference images. Optional upscale (2k) applies to all. |
| generate_videoA | Generate a video with Google Flow (Veo) and save it to disk. UNLIKE images, video COSTS CREDITS: it only runs when the FLOW_MAX_COST ceiling is raised above the quoted cost; otherwise it refuses before generating. Text-to-video from a prompt; pick an aspect (16:9 or 9:16). |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 7 tools
Most tools target clearly distinct actions: status check, single-image generation, download, upscale, library listing, batch, and video. The only mild overlap is between generate_image (which can already produce 'one or more images') and generate_batch (several different prompts), so an agent could occasionally hesitate over which to pick.
Six of seven names follow a clean verb_noun snake_case pattern (generate_image, download_image, upscale_image, list_library, generate_batch, generate_video). The outlier is flow_status, a noun phrase rather than an action, but it remains readable and unambiguous.
Seven tools is well-scoped for a browser-automation image/video generation server, with each tool covering a distinct step in the workflow (diagnose, list, generate, batch, upscale, download, video). Nothing feels padded or missing at the count level.
The lifecycle is largely covered: connection diagnostics, library inspection, generation (single, batch, video), upscaling, and retrieval by media id. The notable gap is there is no tool to upload a new reference image into the Flow library, even though reference images are a first-class concept, forcing the agent back to the UI.