Open Google Image Generator MCP
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| DEFAULT_OUTPUT_DIR | No | Directory where generated images/videos will be saved locally | ./outputs |
| GOOGLE_ACCESS_TOKEN | No | Direct OAuth 2.0 Access Token to bypass Application Default Credentials | |
| GOOGLE_CLOUD_PROJECT | Yes | Your Google Cloud Project ID | |
| GOOGLE_CLOUD_LOCATION | Yes | The Google Cloud region to use (e.g., us-central1, europe-west4) | |
| IMPERSONATE_SERVICE_ACCOUNT | No | Service Account Impersonation for high-security environments |
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": false
} |
| prompts | {
"listChanged": false
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| tool_list_available_modelsA | Live-probe every candidate Vertex AI publisher model in this project/location and return only those that respond (HTTP 200 or 400 = reachable; 404 = not found and excluded). Results are cached for the server process lifetime; pass force_refresh=true to rescan. Returns: { "available": { "image_generation": [...], "image_transformation": [...], "text": [...], "vision": [...] }, "recommended": <static list, may include models that 404 in this project>, "checked_at": , "project": ..., "location": ... } |
| tool_generate_imageA | Generate an image from a text prompt using Vertex AI Gemini image models (Nano Banana). Use model_tier for simple model selection: fast/balanced/quality/ultra. negative_prompt / seed / watermark / safety / storage_uri apply to legacy Imagen models only. |
| tool_edit_imageA | Image editing via Gemini image models — mask-guided inpaint/outpaint, background swap, product image, or mask-free prompt-driven edit. For free-form natural-language transforms (style transfer, scene rewriting), use tool_transform_image instead. |
| tool_transform_imageA | Free-form 'image + text -> image' transformation via Gemini multimodal models. Use for style transfer, scene rewriting, or any natural-language image edit that does not require pixel-precise masking. |
| tool_analyze_imageC | Analyze an image using a multimodal prompt with Vertex AI. |
| tool_upscale_imageB | Upscale a low resolution image to 2K or 4K by re-rendering it with a Gemini image model. |
| tool_remove_backgroundB | Remove the background from an image and save a transparent PNG. The Gemini model places the subject on a flat chroma-key colour, which is then keyed out locally. |
| tool_generate_videoC | Generate a video from a text prompt using Veo. |
| tool_image_to_videoC | Generate a video from a still image as the first frame using Veo. |
| tool_extend_videoC | Extend an existing video by generating additional seconds at the end. |
| tool_video_object_editC | Insert or remove an object in a video using Veo. |
| tool_upload_fileA | Register a local file for use as a reference image in other tools (e.g. tool_transform_image). Returns a file_uri to pass as additional_image_paths. |
| tool_batch_generateA | Generate images for multiple prompts in a single call (max 10, max 4 concurrent). |
| tool_run_pipelineB | Chain image processing steps sequentially. Each step's output becomes the next step's input. Supported tools: generate, edit, transform, upscale, remove_background. Example: generate → remove_background → upscale. |
| tool_generate_musicB | Generate music from a text prompt using Lyria. Check tool_list_available_models to verify Lyria is enabled in your project. |
| tool_embedC | Generate a float embedding vector for the given text using Gemini Embedding. |
| tool_analyze_videoB | Analyze a local video file using Gemini Vision. Supports mp4, mov, avi, webm, mkv. Max 20MB inline. |
| tool_generate_speechB | Convert text to speech using Gemini TTS. Returns a WAV file path in outputs/. |
| tool_live_generateC | Generate text using Gemini streaming. The full streamed response is accumulated and returned. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
| character_design | Template for creating a detailed character design image prompt following Gemini 3 guidelines. |
| logo_concept | Template for designing a modern logo following Gemini 3 guidelines. |
| UI_UX_mockup | Template for generating UI/UX mockups following Gemini 3 guidelines. |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 19 tools
Most tools target clearly distinct resource/action pairs across image, video, audio, and text. However, tool_edit_image and tool_transform_image overlap: edit_image mentions 'mask-free prompt-driven edit' while transform_image is for free-form natural-language transforms, so the boundary between them is slightly blurry despite the descriptions attempting to redirect.
All 19 tools use the same tool_ prefix with snake_case verb_noun or verb_noun_phrase pattern, e.g. tool_generate_image, tool_remove_background, tool_run_pipeline. No mixed conventions or inconsistent verb styles.
19 tools is on the heavy side for a single MCP server, even accounting for the broad multimodal scope. Some tools, such as batch_generate and run_pipeline, are convenience wrappers that an agent could orchestrate by calling individual generation tools, making the surface feel somewhat inflated.
The surface covers generation, editing, transformation, upscaling, background removal, video create/extend/edit/analyze, music, speech, text, embeddings, model discovery, batch, and pipeline. Minor gaps include no explicit output management/history tool and no non-streaming text generation, but core multi-modal workflows are well covered.