GoogleFlow_MCP
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| GEMINI_API_KEY | No | API key de Gemini. Obligatoria si no usas Vertex AI. | |
| GOOGLE_API_KEY | No | Alternativa a GEMINI_API_KEY; también vale como API key de Gemini. | |
| FLOW_OUTPUT_DIR | No | Carpeta donde se guardan videos, imágenes y manifest.json. | ./flow-output |
| FLOW_IMAGE_MODEL | No | Modelo de imagen (Nano Banana Pro). También acepta modelos imagen-*. | gemini-3-pro-image-preview |
| FLOW_VIDEO_MODEL | No | Modelo de video principal. | veo-3.1-generate-preview |
| GOOGLE_CLOUD_PROJECT | No | Proyecto de Google Cloud para Vertex AI. | |
| FLOW_VIDEO_FAST_MODEL | No | Modelo usado con fast: true (borradores más baratos). | veo-3.1-fast-generate-preview |
| GOOGLE_CLOUD_LOCATION | No | Región de Vertex AI. | us-central1 |
| GOOGLE_GENAI_USE_VERTEXAI | No | Usa true para usar Vertex AI. | false |
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
} |
| prompts | {
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| flow_generate_videoA | Create a short video clip with Veo (the model behind Google Flow). Modes: text-to-video (prompt only); frames-to-video (startImage, optional endImage); ingredients-to-video (referenceImages for consistent characters/objects/style). Write a cinematic prompt: subject, action, setting, camera move, lens, lighting, style, and any dialogue in quotes plus sound effects/ambience. Images can be asset ids (from flow_list_assets / flow_generate_image) or local file paths. Rendering takes 1-6 minutes; by default the tool waits. If it returns status=pending, call flow_check_operation later. Files are saved locally and the result includes their paths. |
| flow_extend_videoA | Continue a previously generated Veo video with a new prompt, like Flow's Extend / Scene Builder. The source must be a video asset generated by this server (Gemini API keeps the remote file ~2 days, 720p only). |
| flow_check_operationA | Check (and optionally wait for) a pending Veo operation. When finished, the video is downloaded and its asset id/path returned. Call without operationName to list all pending renders. |
| flow_generate_imageA | Create or edit images with Google's image models (Nano Banana / Imagen), as in Flow's image tools. Without inputImages: text-to-image. With inputImages: edit, restyle, or combine them according to the prompt. Use it to build characters, props and keyframes, then pass the returned asset ids to flow_generate_video as startImage/endImage/referenceImages. Returns local paths and a preview. |
| flow_list_assetsA | List videos and images created by this server (newest first), with ids, prompts and local paths. |
| flow_list_modelsA | List the Veo / image model ids your API key can use, plus the defaults this server is configured with. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
| flow_director | Turn a story idea into a shot list and render it with Veo, step by step. |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 6 tools
Each tool targets a distinct operation: video generation, video extension, image generation, operation status, asset listing, and model listing. The only conceptual proximity is between flow_generate_video and flow_extend_video, but their descriptions clearly separate creation from continuation.
All tools follow a consistent flow_verb_noun pattern in snake_case: flow_generate_video, flow_extend_video, flow_check_operation, flow_generate_image, flow_list_assets, flow_list_models. There are no mixed conventions or vague names.
Six tools are well-scoped for a media-generation server, covering generation, extension, status checking, image creation, and listing assets/models. Each tool earns its place without bloating the surface.
The surface covers the core Flow workflow: generate video/images, extend videos, check pending renders, and list assets/models. Minor lifecycle gaps exist, such as canceling operations or deleting assets, but agents can work around them.